Opinion for the court, in which Chief Judge BAZELON and Circuit Judges McGOWAN, LEVENTHAL, and SPOTTSWOOD W. ROBINSON, III, concur, filed by Circuit Judge J. SKELLY WRIGHT.
Concurring opinion, in which Circuit Judge McGOWAN joins, filed by Chief Judge BAZELON.
Concurring statement filed by Circuit Judge LEVENTHAL.
Dissenting opinion filed by Circuit Judge MacKINNON.
Dissenting opinion, in which Circuit Judges TAMM and ROBB join, filed by Circuit Judge WILKEY.
J. SKELLY WRIGHT, Circuit Judge:Man’s ability to alter his environment has developed far more rapidly than his ability to foresee with certainty the effects of his alterations. It is only recently that we have begun to appreciate the danger posed by unregulated modification of the world around us, and have created watchdog agencies whose task it is to warn us, and protect us, when technological “advances” present dangers unappreciated — or unrevealed — by their supporters. Such agencies, unequipped with crystal balls and unable to read the future, are nonetheless charged with evaluating the effects of unprecedented environmental modifications, often made on a massive scale. Necessarily, they must deal with predictions and uncertainty, with developing evidence, with conflicting evidence, and, sometimes, with little or no evidence at all. Today we address the *379scope of the power delegated one such watchdog, the Environmental Protection Agency (EPA). We must determine the certainty required by the Clean Air Act before EPA may act to protect the health of our populace from the lead particulate emissions of automobiles.
Section 211(c)(1)(A) of the Clean Air Act1 authorizes the Administrator of EPA to regulate gasoline additives whose emission products “will endanger the public health or welfare * * 42 U.S.C. § 1857f-6c(c)(l)(A). Acting pursuant to that power, the Administrator, after notice and comment, determined that the automotive emissions caused by leaded gasoline present “a significant risk of harm” to the public health. Accordingly, he promulgated regulations that reduce, in step-wise fashion, the lead content of leaded gasoline.2 We must decide whether the Administrator properly interpreted the meaning of Section 211(c)(1)(A) and the scope of his power thereunder, and, if so, whether the evidence adduced at the rule-making proceeding supports his final determination. Finding in favor of the Administrator on both grounds, and on all other grounds raised by petitioners, we affirm his determination.
I. THE FACTS, THE STATUTE, THE PROCEEDINGS, AND THE REGULATIONS
Hard on the introduction of the first gasoline-powered automobiles came the discovery that lead “antiknock” compounds, when added to gasoline, dramatically increase the fuel’s octane rating. Increased octane allows for higher compression engines, which operate with greater efficiency. Since 1928 antiknocks have been regularly added to gasoline, and a large industry has developed to supply those compounds. Today, approximately 90 percent of motor gasoline manufactured in the United States contains lead additives, even though most 1975 and 1976 model automobiles are equipped with catalytic converters, which require lead-free gasoline. From the beginning, however, scientists have questioned whether the addition of lead to gasoline, and its consequent diffusion into the atmosphere from the automobile emission, poses a danger to the public health.3 As use of *380automobiles, and emission of lead particulates, has accelerated in the last quarter century, this concern has mounted. The reasons for concern are obvious (and essentially undisputed by petitioners): (1) lead in high concentrations in the body is toxic; (2) lead can be absorbed into the body from the ambient air; and (3) lead particulate emissions from gasoline engines account for approximately 90 percent of the lead in our air. Despite these apparent reasons for concern, hard proof of any danger caused by lead automotive emissions has been hard to come by. Part of the reason for this lies in the multiple sources of human exposure to lead.
Lead is an ubiquitous element. It is found in the land, in the sea, in plants, in animals, and, ultimately, in humans. Traces of lead ranging from 10 to 40 micrograms per 100 grams of blood (10-40 ug/100g)4 are found in everyone, including those living in environments with almost no atmospheric lead. National Academy of Sciences Committee on Biologic Effects of Atmospheric Pollutants, Airborne Lead in Perspective 118 (1972) (hereinafter NAS Report). Despite its universal presence, however, lead serves no known purpose in the human body, and at higher concentrations is toxic, causing anemia, severe intestinal cramps, paralysis of nerves, fatigue, and even death. Clinical symptoms of lead poisoning appear at blood lead levels of 80-100 ug or higher, and symptomatic lead poisoning may appear at levels of 50-60 ug, particularly in the presence of anemia. EPA’s Position on the Health Implications of Airborne Lead (hereinafter Third Health Document) at III — 1, Joint Appendix (hereinafter JA) 54-55.
Human body lead comes from three major sources. In most people, the largest source is the diet. EPA estimates daily dietary lead intake for adults to average 200-300 ug per day, with a range of 100-500 ug a day. Third Health Document at V-2, JA 82. Absorption of dietary lead into the bloodstream is estimated at about 10 percent, although in children absorption may be as high as 50 percent. Thus the average adult adds 20-30 ug of lead to his bloodstream daily from his diet alone. This daily intake, which may be highly variable depending on individual diets, NAS Report at 50, is generally regarded as, for all practical purposes, uncontrollable.5
A second major source of the body’s lead burden, at least among urban children, is regarded as controllable, although effective control may be both difficult and expensive to achieve. Ingestion of lead paint by children with pica (the abnormal ingestion of non-food substances, a relatively common trait in pre-school children, particular ages 1-3) is generally regarded as “the principal environmental source in cases of severe acute lead poisoning in young children.” NAS Report at 140. Lead-based paint was widely used in pre-1940 housing, for both interiors and exteriors, so children living in older housing, particularly in urban ghettos where such paint is both present and peeling, are most susceptible to this form of lead poisoning. Limited control has been achieved in that lead paints are now rarely *381used, and are frequently banned by statute, for interior surfaces. But while some local laws require removal of existing peeling lead paints, and there is federal legislation to aid states and municipalities in such efforts, Lead-Based Paint Poisoning Prevention Act, 42 U.S.C. §§ 4801 et seq. (1970), there is no concentrated national effort at removal, and the danger to children living in dilapidated housing will remain for some time.6
The last remaining major source of lead exposure for humans is the ambient air. This source is easily the most controllable, since approximately 90 percent of lead in the air comes from automobile emissions,7 and can be simply eliminated by removing lead from gasoline.8 While the extent to which such lead actually enters the body is vigorously contested by petitioners and lies at the heart of this appeal, all parties agree that, to some extent at least, airborne lead can be absorbed through the lungs as a person breathes lead-contaminated air and that it can be eaten by children with pica after larger lead particles fall to the ground and mix with dust. Once the lead is in the body, however, its source becomes irrelevant; all lead in the bloodstream, from whatever source, is essentially fungible. Thus so long as there are multiple sources of lead exposure it is virtually impossible to isolate one source and determine its particular effect on the body. The effect of any one source is meaningful only in cumulative terms.
The multiple sources of human exposure to lead explain in part why it has been difficult to pinpoint automobile lead emissions as a danger to public health. Obviously, any danger is caused only by the additive effect of lead emissions on the other, largely uncontrollable, sources of lead. For years the lead antiknock industry has refused to accept the developing evidence that lead emissions contribute significantly to the total human lead body burden. In the Clean Air Act Amendments of 1970, Pub.L. 91-604, December 31, 1970, 84 Stat. 1698-1700, however, Congress finally set up a legal mechanism by which that evidence could be weighed in a more objective tribunal. It gave the newly-created EPA authority to control or prohibit the sale or manufacture of any fuel additive whose emission products “will endanger the public health or welfare * * *.” 42 U.S.C. § 1857f-6c(e)(l)(A) (1970). It is beyond question that the fuel additive Congress had in mind was lead.9
Given this mandate, EPA published on January 31,1971 advance notice of proposed rule-making. The Administrator announced he was considering possible controls on lead additives in gasolines, both because of their possible danger to health and because of their incompatibility with the newly-developed catalytic converter emission control system.10 36 Fed.Reg. 1486 (1971). Proposed regulations were issued a year later, February 23, 1972, supported by a document Health Hazards of Lead11 (hereinafter First Health Docu*382ment), prepared by the EPA scientific staff. Comments were invited for a 90-day period, later reopened for an additional 30 days. 37 Fed.Reg. 11786-11787 (1972). At the same time public hearings were held in Washington, D. C., Dallas, and Los Angeles.
On January 10, 1973 the Administrator, while issuing final regulations requiring availability of some lead-free gasoline to allow implementation of the catalytic converter system, 38 Fed.Reg. 1254; approved in Amoco Oil Co. v. EPA, 163 U.S.App.D.C. 162, 501 F.2d 722 (1974), reproposed the health-based regulations now at issue. 38 Fed.Reg. 1258. The reproposal was supported by a second health document, EPA’s Position on the Health Effects of Airborne Lead (hereinafter Second Health Document), JA 158, and was necessitated by a modification of EPA’s analysis of the health effects of lead emissions. The Agency concluded, after considering the comments received, that it was virtually impossible to identify the precise amount of airborne lead that will endanger public health. Instead, the control strategy would concentrate on evaluating the cumulative effect of airborne lead on total human lead exposure and the significance of that contribution. 38 Fed.Reg. 1258. The reproposed regulations themselves were similar to the original proposal, requiring a phased cutback to 1.25 grams of lead per gallon of leaded gasoline, but, in recognition of the industry’s lead-time problems, pushing the timetable for reduction back one year. The Agency again invited public comment, this time for a 60-day period.
On October 28, 1973, as a result of a motion filed in Natural Resources Defense Council, Inc. v. EPA, D.C.Cir. No. 72-2233, this court ordered EPA to reach within 30 days a final decision on whether lead additives should be regulated for health reasons. EPA published its final health document, entitled EPA’s Position on the Health Implications of Airborne Lead, on November 28,1973. JA 27. This document, the Third Health Document, extensively details and reviews the state of knowledge of the health effects of airborne lead. It candidly discusses the various scientific studies, both pro and con, underlying this information, and ultimately concludes that lead from automobile emissions will endanger the public health. The same day, based largely on the conclusions of the Third Health Document, EPA promulgated its final regulations, accompanied by a thorough discussion of its health conclusions, the impact of the regulations, and the alternative courses of action considered and rejected. 38 Fed. Reg. 33734. The final regulations require the same step-wise reduction of lead additives but, in response to the comments of a majority of refiners, calculate the reduction in a slightly different manner. Whereas the original and reproposed regulations set standards for permissible lead use by each refiner on the basis of grams of lead per gallon of leaded gasoline produced (leaded pool averaging), the final regulations base the standards on grams of lead per gallon of all gasoline produced (total pool averaging). The quantity of lead emitted into the atmosphere is the same under both systems; EPA simply converted its leaded pool figures into total pool figures. Under the final regulations, lead in all gasoline would be reduced over a five-year period to an average of 0.5 grams per gallon.12
Petitioners, various manufacturers of lead additives and refiners of gasoline, appealed the promulgation of low-lead regulations to this court under Section 307 of the Clean Air Act, 42 U.S.C. § 1857h-5. The *383appeal was heard by a division of the court on September 9, 1974. On December 20, 1974, the division, one judge dissenting, ordered the regulations set aside. The majority and dissenting opinions were published on January 28, 1975.13 Because of the importance of the issues presented, we granted EPA’s petition for rehearing en banc on March 17, 1975, vacating the judgment and opinions of the division and setting the case for reargument on May 30, 1975. All parties were invited to submit supplementary briefs addressing the issues raised by the division opinions.
The regulations are challenged by petitioners on a variety of grounds, all of which will be addressed below. Their primary claims, and the ones on which the division majority based its reversal, are that the Administrator misinterpreted the statutory standard of “will endanger” and that his application of that standard is without support in the evidence and arbitrary and capricious.
II. THE STATUTORY REQUIREMENTS
Under Section 211(c)(1)(A) the Administrator may, on the basis of all the information available to him, promulgate regulations that
control or prohibit the manufacture, introduction into commerce, offering for sale, or sale of any fuel or fuel additive for use in a motor vehicle or motor vehicle engine (A) if any emission products of such fuel or fuel additive will endanger the public health or welfare * * *.
42 U.S.C. § 1857f-6c(c)(l)(A). The Administrator cannot act under Section 211(c)(1)(A), however, until after “consideration of all relevant medical and scientific evidence available to him, including consideration of other technologically or economically feasible means of achieving emission standards under [Section 20'2].” Section 211(c)(2)(A), 42 U.S.C. § 1857f-6c(c)(2)(A). Section 202 of the Act, 42 U.S.C. § 1857Í-1, allows the Administrator to set standards for emission of pollutants from automobiles (as opposed to standards for the composition of the gasoline that produces the emissions), and is thus the preferred — although not the mandatory — alternative under the statutory scheme, presumably because it minimizes Agency interference with manufacturer prerogatives.14
The Administrator is also required, before prohibiting a fuel or fuel additive under Section 211(c)(1)(A), to find, and publish the finding, that in his judgment any fuel or fuel additive likely to replace the prohibited one will not “endanger the public health or welfare to the same or greater degree * * Section 211(c)(2)(C), 42 U.S.C. § 1857f-6c(c)(2)(C). It is significant *384that this is the only conclusion the Administrator is expressly required to “find” before regulating a fuel or fuel additive for health reasons.
A. The Threshold Determination
In making his threshold determination that lead particulate emissions from motor vehicles “will endanger the public health or welfare,” the Administrator provided his interpretation of the statutory language by couching his conclusion in these words: such emissions “present a significant risk of harm to the health of urban populations, particularly to the health of city children.” 38 Fed.Reg. 33734. By way of further interpretation, he added that it was his view
that the statutory language * * * does not require a determination that automobile emissions alone create the endangerment on which controls may be based. Rather, the Administrator believes that in providing this authority, the Congress was aware that the public’s exposure to harmful substances results from a number of sources which may have varying degrees of susceptibility to control.
Id. It is petitioners’ first claim of error that the Administrator has erroneously in-
terpreted Section 211(c)(1)(A) by not sufficiently appreciating the rigor demanded by Congress in establishing the “will endanger” standard. Therefore, petitioners argue, the Administrator’s action is “short of statutory right,” in violation of Section 10(e)(2)(C) of the Administrative Procedure Act (APA), 5 U.S.C. § 706(2)(C) (1970).
Petitioners argue that the “will endanger” standard requires a high quantum of factual proof, proof of actual harm rather than of a “significant risk of harm.” See supplemental brief of petitioner Ethyl Corporation (hereinafter Ethyl Supp.Br.) at 20. Since, according to petitioners, regulation under Section 211(c)(1)(A) must be premised upon factual proof of actual harm, the Administrator has, in their view, no power to assess risks or make policy judgments in deciding to regulate lead additives. Moreover, petitioners argue, regulation must be based on the danger presented by lead additives “in and of themselves,” so it is improper to consider, as the Administrator did, the cumulative impact of lead additives on all other sources of human exposure to lead. We have considered these arguments with care and find them to be without merit.15 It is our view that the Administrator’s interpretation of the standard is the correct one.16
*3851. The Precautionary Nature of “Will Endanger.” Simply as a matter of plain meaning, we have difficulty crediting petitioners’ reading of the “will endanger” standard. The meaning of “endanger” is not disputed. Case law and dictionary definition agree that endanger means something less than actual harm.17 When one is endangered, harm is threatened; no actual injury need ever occur. Thus, for example, a town may be “endangered” by a threatening plague or hurricane and yet emerge from the danger completely unscathed.18 A statute allowing for regulation in the face of danger is, necessarily, a precautionary statute. Regulatory action may be taken before the threatened harm occurs; indeed, the very existence of such precautionary legislation would seem to demand that regulatory action precede, and, optimally, prevent, the perceived threat. As should be apparent, the “will endanger” language of Section 211(c)(1)(A) makes it such a precautionary statute.
The Administrator read it as such, interpreting “will endanger” to mean “presents a significant risk of harm.” 38 Fed.Reg. 33734. We agree with the Administrator’s interpretation. This conclusion is reached not only by reference to the plain meaning of the statute, but by juxtaposition of Section 211(c)(1)(A) with other sections of the Clean Air Act and by analysis of pertinent precedent.19
Petitioners support their view of the rigorous nature of the “will endanger” standard by relying on two other sections of the Clean Air Act that also allow for regulation of air pollutants for health reasons. Ethyl *386Supp.Br. at 18-23; supplemental brief of petitioner Nalco Chemical Company (hereinafter Nalco Supp.Br.) at 20-25; NPRA Supp.Br. at 31-46. However, we find in the same sections relied upon by petitioners firm support for our view of the precautionary nature of Section 211(c)(1)(A). The provisions identified by petitioners are Sections 108 and 202 of the Act. Section 108 directs the Administrator to list, for the purpose of establishing national primary and secondary ambient air quality standards pursuant to Section 109, each air pollutant “which in his judgment has an adverse effect on public health or welfare * * *.” Section 108(a)(1)(A), 42 U.S.C. § 1857c-3(a)(l)(A). Section 202 authorizes the Administrator to set standards for each automobile emission “which in his judgment causes or contributes to, or is likely to cause or to contribute to, air pollution which endangers the public health or welfare.” Section 202(a)(1), 42 U.S.C. § 1857f-l(a)(l).
Petitioners suggest that the threshold decision to regulate under both Section 108 and Section 202 can be based on less evidence than required under Section 211 and that, therefore, the proof necessary for action under Section 211 must be particularly firm. Petitioners misread both sections. Section 108 requires an actual “adverse effect” on health before an air pollutant may be listed for regulation; actual harm must result.20 This firm threshold finding is reasonable in light of the considerable disruption caused by action under Sections 108-110. After a pollutant is listed under Section 108, the Administrator must issue national ambient air quality standards under Section 109 within a year. Promulgation of standards begins the implementation plan process of Section 110 under which the states must control, on a mandatory timetable, the listed pollutants to the extent necessary to achieve the federal standards. Section 110, 42 U.S.C. § 1857c-5. Sections 108-110 are “technology forcing” provisions; the attainment of the primary, health-based standards takes precedence over the cost and present technological feasibility of achieving the requisite control. Union Electric Co. v. EPA, 515 F.2d 206, 215-16 (8th Cir.), cert. granted, 423 U.S. 821, 96 S.Ct. 35, 46 L.Ed.2d 38, 44 U.S.L. Week 3200 (1975). Cf. Train v. Natural Resources Defense Council, 421 U.S. 60, 90-91, 95 S.Ct. 1470, 1487-1488, 43 L.Ed.2d 731, 752-753 (1975). See also S.Rep.No.91-1196, 91st Cong., 2d Sess. 1-3 (1970).
Thus, before ordering this extensive chain of action to begin, Congress demanded a threshold determination that the pollutant causes actual harm.21 In this sense Section 108 is not a precautionary statute at all, and so differs sharply from Section 211. However, the effects of such after-the-fact regulation are somewhat ameliorated by the Act; Congress did provide a precautionary element in standard-setting under Sections 108-110. Section 109 expressly requires that the ambient air standards ultimately issued provide for “an adequate-*387margin of safety.” 42 U.S.C. § 1857c-4(b)(1). Thus, while the threshold decision to regulate under Sections 108-110 is not precautionary but rather requires proof of demonstrable harm caused by the suspect pollutant, once the decision is made the standards promulgated must be preventive in nature. Congress’ choice of this scheme is in direct contrast to the procedures it established under Section 211. Under that section the decision to regulate is based on perceived danger. Unless we are to assume Congress chose its language carelessly, regulation in the face of “danger” rather than in the face of “adverse effects” must mean that the threshold decision to regulate under Section 211 is precautionary.22 The contrast between the standards of Sections 108-110 and of Section 211 supports our view of the precautionary nature of the “will endanger” standard.23
Petitioners also rely on Section 202 to support their strict reading of Section 211. Ethyl suggests that Section 202 is more lenient than Section 211 in that it allows regulation of “likely” dangers. Ethyl Supp.Br. at 18-19. See also Nalco Supp.Br. at 20-21; NPRA Supp.Br. at 31-36. Section 202 provides that the Administrator may regulate
The emission of any air pollutant [from any new motor vehicle] which in his judgment causes or contributes to, or is likely to cause or to contribute to, air pollution which endangers the public health or welfare.
42 U.S.C. § 1857f — 1(a)(1) (emphasis added). While this language may be unnecessarily opaque, we think a fair reading disproves petitioners’ suggestion. The italicized language upon which petitioners rely refers not to the causal relationship between air pollution and health, but to the relationship between automobile emissions and air pollution.24 Thus regulation may not be premised on a threshold determination of likely danger; rather regulation must be premised on a determination of danger, a finding that “air pollution which endangers the public health” is the end product of the emission to be regulated. This is essentially the same finding of endangerment as under *388Section 211. “Likely”, enters the equation only in determining whether the emitted air pollutant, which would be regulated, contributes to the air pollution which is found dangerous. Here the statute allows for a somewhat attenuated chain of causation. Regulation may be premised on a determination that an air pollutant emitted from a new automobile is likely to contribute to air pollution which endangers the public health. In establishing this chain of causation Section 202 is more lenient than Section 211,25 but in making the threshold determination of danger both sections are the same:26 air pollution must endanger the public health before regulation is justified.27
Thus the two sections of the Clean Air Act presented by petitioners in support of their stringent reading of the threshold requirement for action under Section 211 turn out, upon analysis, to be of no support at all. Section 108 only bolsters our reading of Section 211 as a precautionary statute while Section 202 includes the same standard as Section 211 and thus is of no guidance one way or the other. While petitioners have little more to offer to prove that the “will endanger” standard demands proof of actual harm and is not precautionary in nature, we may turn, in support of our interpretation, to the relevant case law. While cases interpreting the meaning of “endanger” are few in number,28 at least *389one recent case is directly on point and fully in accord with our view.
In Reserve Mining Co. v. EPA, 514 F.2d 492 (8th Cir. 1975) (en banc), the Eighth Circuit addressed, among other issues, the meaning of the phrase “endangering the health or welfare of persons” under Section 112 of the Federal Water Pollution Control Act of 1970 (FWPCA), 33 U.S.C. § 1160. FWPCA and the Clean Air Act together constitute the bulk of this nation’s substantive environmental protection legislation.29 As such, and because of their contemporaneous enactment, interpretations of provisions of one Act have frequently been applied to comparable provisions of the other. See, e. g., Natural Resources Defense Council, Inc. v. Train, 166 U.S.App.D.C. 312, 321-322, 510 F.2d 692, 701-702 (1975). Thus Reserve Mining’s interpretation of “endangering” is relevant to the meaning of the term “endanger” in the Clean Air Act. Indeed, it is particularly relevant because in construing the language before it the Eighth Circuit borrowed extensively from the interpretation of the “will endanger” language of Section 211 expressed in the dissent from the division opinion in this case, the same interpretation we adopt here. See Reserve Mining Co. v. EPA, supra, 514 F.2d at 528-529. After analysis of the plain meaning of the FWPCA provision, comparison with other sections of that Act, and reference to our division’s dissent, the Eighth Circuit’s unanimous conclusion fully supports our view of the “will endanger” standard:
In the context of this environmental legislation, we believe that Congress used the term “endangering” in a precautionary or preventive sense, and, therefore, evidence of potential harm as well as actual harm comes within the purview of that term.
Id. at 528.
In sum, based on the plain meaning of the statute, the juxtaposition of Section 211 with Sections 108 and 202, and the Reserve Mining precedent, we conclude that the “will endanger” standard is precautionary in nature and does not require proof of actual harm before regulation is appropriate.30
Perhaps because it realized that the above interpretation was the only possible reading of the statutory language, petitioner Ethyl addresses this interpretation and argues that even if actual harm is not re*390quired for action under Section 211(c)(1)(A), the occurrence of the threatened harm must be “probable” before regulation is justified. Ethyl Supp.Br. 12. While the dictionary admittedly settles on “probable” as its measure of danger, we believe a more sophisticated case-by-case analysis is appropriate.31 Danger, the Administrator recognized, is not set by a fixed probability of harm, but rather is composed of reciprocal elements of risk and harm, or probability and severity. Cf. Carolina Environmental Study Group v. United States, 166 U.S.App.D.C. 416, 419, 510 F.2d 796, 799 (1975); Reserve Mining Co. v. EPA, supra, 514 F.2d at 519-520. That is to say, the public health may properly be found endangered both by a lesser risk of a greater harm and by a greater risk of a lesser harm.32 Danger depends upon the relation between the risk and harm presented by each case, and cannot legitimately be pegged to “probable” harm, regardless of whether that harm be great or small. As the Eighth Circuit found in Reserve Mining, these concepts “necessarily must apply in a determination of whether any relief should be given in cases of this kind in which proof with certainty is impossible.”33 514 F.2d at 520.
In Reserve Mining the issue was whether asbestiform wastes flushed into Lake Superior by the Reserve Mining Company endangered health.34 The polluted lake *391waters formed the drinking supply of several surrounding communities, while a medical theory, bolstered only by inconclusive evidence, suggested that ingestion of the wastes caused cancer. See pages 417-418 of 176 U.S.App.D.C., pages 45-46 of 541 F.2d infra. Applying the “endangering the health or welfare of persons” standard of the FWPCA, the court found the wastes to be a danger cognizable under the Act. The court did not find that the danger was probable; rather it found the wastes to be “potentially harmful,” 514 F.2d at 528, and potential harm to be embraced by the “endangering” standard, id. See page 389 of 176 U.S.App.D.C., page 17 of 541 F.2d supra. The court concluded:
The record shows that Reserve is discharging a substance into Lake Superior waters which under an acceptable but unproved medical theory may be considered as carcinogenic. As previously discussed, this discharge gives rise to a reasonable medical concern over the public health. We sustain the district court’s determination that Reserve’s discharge into Lake Superior constitutes pollution of waters “endangering the health or welfare of persons” within the terms of §§ 1160(c)(5) and (g)(1) of the Federal Water Pollution Control Act and is subject to abatement.
514 F.2d at 529 (footnote omitted; emphasis added). The court thus allowed regulation of the effluent on only a “reasonable” or “potential” showing of danger, hardly the “probable” finding urged by Ethyl as the proper reading of the “endanger” language in Section 211. The reason this relatively slight showing of probability of risk justified regulation is clear: the harm to be avoided, cancer, was particularly great. However, because the risk was somewhat remote, the court did not order the immediate cessation of asbestiform dumping, but rather ordered such cessation within “a reasonable time.” Id. at 538.
Reserve Mining convincingly demonstrates that the magnitude of risk sufficient to justify regulation is inversely proportional to the harm to be avoided. Cf. Carolina Environmental Study Group v. United States, supra. It would be a bizarre exercise in balancing horrors to determine whether cancer or lead poisoning is a greater harm to be avoided,35 but fortunately such balancing is unnecessary in this case. Undoubtedly, the harm caused by lead poisoning is severe; nonetheless, the Administrator does not rely on a “potential” risk or *392a “reasonable medical concern” to justify the regulations before us. Instead, he finds a “significant” risk of harm to health. While this finding may be less than the “probable” standard urged by Ethyl, it is considerably more certain than the risk that justified regulation in Reserve Mining of a comparably “fright-laden” harm. Cf. Environmental Defense Fund, Inc. v. EPA, 150 U.S.App.D.C. 348, 358, 465 F.2d 528, 538 (1972). Moreover, like the Reserve Mining court, in the face of this still less than certain risk the Administrator did not order the cessation of use of lead additives, but rather directed a phased step-down to a plateau level. Thus we conclude that however far the parameters of risk and harm inherent in the “will endanger” standard might reach in an appropriate case, they certainly present a “danger” that can be regulated when the harm to be avoided is widespread lead poisoning and the risk of that occurrence is “significant.”36
2. The Administrator’s Power to Assess Risks. Petitioners argue that Section 211 requires the Administrator to make a “threshold factual determination” that automobile emissions “will endanger” the public health, Nalco Supp.Br. at 15-20; Ethyl Supp.Br. at 24-26, and dispute EPA’s claim that the Administrator may make “an essentially legislative policy judgment, rather than a factual determination, concerning the relative risks of underprotection as compared to overprotection.” Industrial Union Department, AFL-CIO v. Hodgson, 162 U.S.App.D.C. 331, 339, 499 F.2d 467, 475 (1974). We must reject .petitioners’ argument, since the power to assess risks, without relying solely on facts, flows inexorably from the nature of the “will danger” standard. We have already found that Section 211 allows the Administrator to regulate fuel content when he finds that emissions cause a significant risk of harm to the public health. Yet, how can the Administrator determine that a risk is a significant risk if he cannot assess risks? And how can he assess risks if he cannot make policy judgments? Surely reliance on “facts” as contemplated by petitioners will provide little guidance. However, sole reliance on facts was not demanded by Congress.37
*393Originally, it is true, it appeared that Congress would severely restrict the Administrator’s ability to assess risks and make policy judgments to protect public health. The bill sent to the floor of the House, and eventually passed by the House, would have allowed EPA to control fuels or fuel additives only if the determination that their emission products would endanger the public health were established
on the basis of specific findings derived from relevant medical and scientific evidence, including * * * a finding that it is not otherwise technologically or economically feasible to achieve the emission standards established pursuant to section 202 of this Act.
H.R. 17255, 91st Cong., 2d Sess. § 210(g)(1) (1970) (emphasis added). The specific findings requirement had more than procedural significance. It limited the scope of evidence on which the Administrator could proceed, made alternative action under Section 202 mandatory if possible, and was widely interpreted in the House, as the extracts of debate relied upon by petitioners show, Ethyl Supp.Br. at 24; NPRA Supp.Br. at 30; Nalco Supp.Br. at 15, as requiring that any EPA action be based solely on facts.38 See also H.R.Rep. No. *39491-1146, 91st Cong., 2d Sess., at 13 (1970), U.S.Code Cong. & Admin.News 1970, p. 5356.
But the House bill did not become law. The Senate’s preference for less restriction of EPA freedom in regulating fuel addifives for health reasons39 was adopted by the conference committee and ultimately enacted into law.40 Although the legislative history does not expressly discuss the reasons for the change,41 the contrast in language is stark. As Section 211 now *395reads, the only “finding” the Administrator is required to make is that any fuel additive that might replace one that is prohibited does not cause emissions that will pose the same or greater danger to the public health.42 The substantive impact of the change is clear. All of the requirements for specific findings quoted above are replaced by requirements that the Administrator “consider” the specified evidence. The mandatory deference to Section 202 is removed.43 And the Administrator may act based on all information available to him.44 As we recognized in Amoco, construing a similar change in parallel Section 211(c)(1)(B), the conference committee’s decision “was a deliberated one and was meant to have significance.” Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 173, 501 F.2d at 733.
In this case the decision has even more significance than it did in Amoco, for under Section 211(c)(1)(B), at issue in Amoco, the Administrator was still required to make “findings.” All the conferees dropped was the adjective “specific.” For regulation under Section 211(c)(1)(A), at issue here, however, the entire requirement was dropped, and no greater restriction was placed on the Administrator than that required by the “basis and purpose” statement of Section 4(b) of the APA, 5 U.S.C. § 553(c). We interpreted the meaning of the limited findings requirement for acting under Section 211(c)(1)(B) in Amoco:
[W]e read Section 211(c)(2)(B) as incorporating the commonsense approach which the courts have developed in applying Section 4(b) [45] of the APA. Where EPA’s regulations turn crucially on factual issues, we will demand sufficient attention to these in the statement to allow the fundamental rationality of the regulations to be ascertained. Where, by contrast, the regulations turn on choices of policy, on an assessment of risks, or on predictions dealing with matters on the frontiers of scientific knowledge, we will demand adequate reasons and explanations, but not “findings” of the sort familiar from the world of adjudication.
Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 180-181, 501 F.2d at 740-741 (emphasis added). Thus the Amoco court read the limited findings necessary for action under Section 211(c)(1)(B) as a flexible requirement that demanded actual findings for questions of fact and something less— “adequate reasons and explanations” — for questions of policy.
As petitioners correctly point out, the Amoco court applied this approach by holding that the threshold determination whether to regulate at all under Section 211(c)(1)(B) was essentially a question of fact, for which a factual finding must be offered. In suggesting that such a finding is necessary for action under Section 211(c)(1)(A) as well, however, petitioners completely ignore the fact that a finding is required for action under Section 211(c)(1)(B) and is not required under Section 211(c)(1)(A). By so glossing over the language of the statute, petitioners miss a vital difference between the threshold determinations under Sections 211(c)(1)(A) and (B). Recall what the Administrator must find in order to act under Section *396211(c)(1)(B): that the emission products of the fuel or additive to be regulated
will impair to a significant degree the performance of any emission control device or system which is in general use, or which the Administrator finds has been developed to a point where in a reasonable time it would be in general use were such regulation to be promulgated.
42 U.S.C. § 1857f-6c(c)(l)(B). At its core, this is a peculiarly factual finding. Will the emission impair the performance of a control device? Is the device in general use, or has it the potential to be? These were the questions, highly suitable to factual proof, that the Amoco court construed. Contrary to petitioners’ implication, the court did not hold that any threshold determination to regulate must be based on factual proof, but that this threshold determination must be factually based. The Section 211(c)(1)(A) threshold determination, however, is inherently unlike that of its sister section. “Endanger,” as we have suggested above, is not a standard prone to factual proof alone. Danger is a risk, and so must be decided by assessment of risks as well as by proof of facts.
Thus a reason emerges for the deletion of the findings requirement for action under the “will endanger” standard. The Amoco court held the findings requirement of Section 211(c)(1)(B) no more than reiterated the minimal demands of the basis and purpose statement of the APA. Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 179, 501 F.2d at 739. What Congress was doing, then, was using the findings requirement to indicate which section demanded a factual threshold determination and which section did not. While inclusion of the phrase was unnecessary in that the APA would have demanded a factual threshold determination under Section 211(c)(1)(B) in any case, the selective use of the findings requirement served to emphasize to the Administrator the demands of the APA and the intent of the Congress. While Congress did not discuss the extent of the Administrator’s power under the “will endanger” standard, its actions — the statute it enacted and the one it rejected — make the legislative intent clear. We find that deletion of the findings requirement for action under Section 211(c)(1)(A) was a recognition by Congress that a determination of endangerment to public health is necessarily a question of policy that is to be based on an assessment of risks and that should not be bound by either the procedural or the substantive rigor proper for questions of fact.
This conclusion follows not only from the language of Section 211(c)(1)(A) and its legislative history, but from the nature of the Administrator’s charge: to protect the public from danger. Regulators such as the Administrator must be accorded flexibility, a flexibility that recognizes the special judicial interest in favor of protection of the health and welfare of people, even in areas where certainty does not exist. Environmental Defense Fund, Inc. v. Ruckelshaus, 142 U.S.App.D.C. 74, 88, 439 F.2d 584, 598 (1971).
Questions involving the environment are particularly prone to uncertainty. Technological man has altered his world in ways never before experienced or anticipated. The health effects of such alterations are often unknown, sometimes unknowable. While a concerned Congress has passed legislation providing for protection of the public health against gross environmental modifications,46 the regulators entrusted with the enforcement of such laws have not thereby been endowed with a prescience that removes all doubt from their decision-making. Rather, speculation, conflicts in evidence, and theoretical extrapolation typify their every action. How else can they act, given a mandate to protect the public health but only a slight or nonexistent data base upon which to draw? Never before have massive quantities of asbestiform tailings been spewed into the water we drink.47 *397Never before have our industrial workers been occupationally exposed to vinyl chloride48 or to asbestos dust.49 Never before has the food we eat been permeated with DDT50 or the pesticides aldrin and dieldrin.51 And never before have hundreds of thousands of tons of lead emissions been disgorged annually into the air we breathe. Sometimes, of course, relatively certain proof of danger or harm from such modifications can be readily found. But, more commonly, “reasonable medical concerns” and theory long precede certainty. Yet the statutes — and common sense — demand regulatory action to prevent harm, even if the regulator is less than certain that harm is otherwise inevitable.
Undoubtedly, certainty is the scientific ideal — to the extent that even science can be certain of its truth.52 But certainty in the complexities of environmental medicine may be achievable only after the fact, when scientists have the opportunity for leisurely and isolated scrutiny of an entire mechanism. Awaiting certainty will often allow for only reactive, not preventive, regulation.53 Petitioners suggest that anything less than certainty, that any speculation, is irresponsible. But when statutes seek to avoid environmental catastrophe, can preventive, albeit uncertain, decisions legitimately be so labeled?
The problems faced by EPA in deciding whether lead automotive emissions pose a threat to the public health highlight the limitations of awaiting certainty. First, lead concentrations are, even to date, essentially low-level, so that the feared adverse effects would not materialize until after a lifetime of exposure. Contrary to petitioners’ suggestion, however, we have not yet suffered a lifetime of exposure to lead emissions. At best, emissions at present levels have been with us for no more than 15-20 years.54 Second, lead exposure from the ambient air is pervasive, so that valid control groups cannot be found against which the effects of lead on our population can be measured. Third, the sources of human exposure to lead are multiple, so *398that it is difficult to isolate the effect of automobile emissions. Lastly, significant exposure to lead is toxic, so that considerations of decency and morality limit the flexibility of experiments on humans that would otherwise accelerate lead exposure from years to months, and measure those results.55 Cf. Environmental Defense Fund, Inc. v. EPA (Shell), 167 U.S.App.D.C. 71, 78, 510 F.2d 1292, 1299 (1975).
The scientific techniques for attempting to overcome these limitations are several: toxicology can study the distribution and effect of lead in animals; epidemiological techniques can analyze the effects of lead emissions on entire populations; clinical studies can reproduce in laboratories atmospheric conditions and measure under controlled circumstances the effects on humans. All of these studies are of limited usefulness, however. Dr. J. H. Knelson, Director of EPA’s Human Studies Laboratory, has described, in the context of setting ambient air standards, the limitations of these various investigative tools:
Each of these investigative approaches — classic toxicology, epidemiology, and clinical research has its advantages and disadvantages. The toxicologist can control the dose and use invasive or destructive techniques in measuring response in the animal, but is always faced with the problem of extrapolating results to humans. Epidemiology is most relevant because it studies phenomena actually occurring in humans under “natural” conditions, but can only draw inference from observed correlations rather than prove cause and effect relationships. Clinical research can provide the most accurate dose-response relationships in the species of interest. Precisely because the study subjects are humans, however, many experimental design problems are encountered in assuring their safety. Although the dose of an atmospheric pollutant can be carefully controlled and measured in the clinical laboratory, qualitative comparability to the multiplex variable of atmospheric pollution cannot always be assured.
The best scientific criteria for establishing air quality standards result from interactions between these disciplines. Clinical studies must be preceded by exhaustive toxicological assessment in other species; observations from population studies should play an important role in the experimental design of clinical research. Biomedical data from all these sources, taken in their entirety, should be used for the prudent definition of air pollution control needs.
JA 582-583. The best biomedical evidence will be derived from relating all three research approaches. This EPA did. That petitioners, and their scientists, find a basis to disagree is hardly surprising, since the results are still uncertain, and will be for some time. But if the statute accords the regulator flexibility to assess risks and make essentially legislative policy judgments, as we believe it does, preventive regulation based on conflicting and inconclusive evidence may be sustained. Recent cases have recognized this flexibility in similar situations.
In Industrial Union Department, AFL-CIO v. Hodgson, supra, this court considered the Secretary of Labor’s delegated power under the Occupational Safety and Health Act (OSHA), 29 U.S.C. § 651 et seq., to protect the health of industrial workers by setting standards for exposure to industrial pollutants. Under review was a standard for exposure to asbestos dust, thought to be carcinogenic. Judge McGowan, writing for the division, laid down the rule:
From extensive and often conflicting evidence, the Secretary in this case made numerous factual determinations. With respect to some of those questions, the evidence was such that the task consisted primarily of evaluating the data and drawing conclusions from it. The court *399can review that data in the record and determine whether it reflects substantial support for the Secretary’s findings. But some of the questions involved in the promulgation of these standards are on the frontiers of scientific knowledge, and consequently as to them insufficient data is presently available to make a fully informed factual determination. Decision making must in that circumstance depend to a greater extent upon policy judgments and less upon purely factual analysis.18
162 U.S.App.D.C. at 338, 499 F.2d at 474.
This rule was likewise applied in Society of the Plastics Industry, Inc. v. OSHA, 509 F.2d 1301 (2d Cir.), cert. denied, 421 U.S. 992, 95 S.Ct. 1998, 44 L.Ed.2d 482 (1975), where the Second Circuit reviewed regulations limiting industrial exposure to vinyl chloride, also considered a carcinogen:
As in Industrial Union Department, AFL-CIO v. Hodgson, supra, the ultimate facts here in dispute are “on the frontiers of scientific knowledge,” and, though the factual finger points, it does not conclude. Under the command of OSHA, it remains the duty of the Secretary to act to protect the workingman, and to act even in circumstances where existing methodology or research is deficient. The Secretary, in extrapolating the MCA study’s finding from mouse to man, has chosen to reduce the permissible level to the lowest detectable one. We find no error in this respect.
Id. at 1308. And in Reserve Mining the Eighth Circuit agreed. Although reviewing the determination of a District Court, not an expert Administrator, the court recognized that it must nonetheless apply the test of the preventive statute before it:
[W]e note that many of the issues in this case do not involve “historical” facts subject to the ordinary means of judicial resolution. Indeed, a number of the disputes involve conflicting theories and experimental results, about which it would be judicially presumptuous to offer conclusive findings [quoting Amoco Oil Co. v. EPA, 163 U.S.App.D.C. 181, 501 F.2d at 741, which relied on Industrial Union]. In such circumstances, the finder of fact must accept certain areas of uncertainty, and the findings themselves cannot extend further than attempting to assess or characterize the strengths and weaknesses of the opposing arguments.
514 F.2d at 507 n.20. See also id. at 529.
These cases, recognizing as they do the developing nature of environmental medicine, fortify our analysis of the “will endanger” language of Section 211.56 *400Where a statute is precautionary in nature,57 the evidence difficult to come by, uncertain, or conflicting because it is on the frontiers of scientific knowledge, the regulations designed to protect the public health, and the decision that of an expert administrator, we will not demand rigorous step-by-step proof of cause and effect. Such proof may be impossible to obtain if the precautionary purpose of the statute is to be served. Of course, we are not suggesting that the Administrator has the power to act on hunches or wild guesses. Amoco makes it quite clear that his conclusions must be rationally justified. Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 180-181, 501 F.2d at 740-741. However, we do hold that in such cases the Administrator may assess risks. He must take account of available facts, of course, but his inquiry does not end there. The Administrator may apply his expertise to draw conclusions from suspected, but not completely substantiated, relationships between facts, from trends among facts, from theoretical projections from imperfect data, from probative preliminary data not yet certifiable as “fact,” and the like. We believe that a conclusion so drawn — a risk assessment— may, if rational, form the basis for health-related regulations under the “will endanger” language of Section 211.58
*401All of this is not to say that Congress left the Administrator free to set policy on his own terms. To the contrary, the policy guidelines are largely set, both in the statutory term “will endanger” and in the relationship of that term to other sections of the Clean Air Act. These prescriptions direct the Administrator’s actions. Operating within the prescribed guidelines, he must consider all the information available to him. Some of the information will be factual, but much of it will be more speculative — scientific estimates and “guesstimates” of probable harm, hypotheses based on still-developing data, etc. Ultimately he must act, in part on “factual issues,” but largely “on choices of policy, on an assessment of risks, [and] on predictions dealing with matters on the frontiers of scientific knowledge * * Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 181, 501 F.2d at 741. A standard of danger — fear of uncertain or unknown harm — contemplates no more.
3. Propriety of the Cumulative Impact Approach. In addition to demanding that the Administrator act solely on facts, petitioner Ethyl insists that those facts convince him that the emission product of the additive to be regulated “in and of itself,” i. e., considered in isolation, endangers health. The Administrator contends that the impact of lead emissions is properly considered together with all other human exposure to lead. See page 384 of 176 U.S.App.D.C., page 12 of 541 F.2d supra. We agree.
First, Ethyl points to the language of Section 202 allowing regulation of any automobile emission that “causes or contributes to, or is likely to cause or to contribute to, air pollution which endangers the public health or welfare.” Section 202(a)(1), 42 U.S.C. § 1857f-l(a)(l) (emphasis Ethyl’s in its brief at 18). This is contrasted with Section 211’s language allowing regulation of fuels or fuel additives whose “emission products * * * will endanger the public health or welfare.” By italicizing the “contribute to” language of Section 202 Ethyl presumably suggests that Section 202 is more lenient, allowing use of the cumulative impact theory while Section 211 does not. This argument is but a variant on Ethyl’s already rejected claim that Section 202 allows for regulation of “likely” dangers while Section 211 does not. As with the earlier argument, this reading of Section 202 must be rejected. While it is possible that Section 202’s inclusion of emissions that “cause or contribute to air pollution” is more encompassing than Section 211’s term “emission products,” it is clear that whatever leniency Section 202 suggests only concerns the makeup of air pollution. This is irrelevant to the cumulative impact theory. However they differ, both Section 202 and Section 211 allow regulation of lead additives only on a determination that the resultant polluted air, however composed, endangers the public health.59 Both provisions leave open the question whether emissions or air pollution can be found to endanger the public health when the endangerment is not caused by that pollution alone.60
While Ethyl’s comparison suggests no answer, the question is directly raised here. The Administrator found endangerment, but recognized that the national lead exposure problem is caused, not by air pollution alone, but by an aggregate of sources, including food, water, leaded paint, and dust. *402He believed that regulation was justified because the aggregate was dangerous, and because leaded gasoline was a significant source that was particularly suited to ready reduction. 38 Fed.Reg. 33734. To the question whether the Administrator was correct in his belief, comparison with the “contribute to” language of Section 202 provides no guidance.
The only other evidence relied upon by Ethyl for its “in and of itself” theory is a quotation, taken out of context, from Senator Baker. When the context of Ethyl’s partial quotation is revealed the irrelevance of his statement to Ethyl’s theory becomes clear. Senator Baker reviewed the Committee bill for the Senate:
The [Administrator] is authorized to either control or prohibit the sale of any given fuel when he finds one of two things:
First. That the combustion or evaporation of such fuel produces emissions that, in and of themselves, endanger the public health or welfare; or Second. That such emissions prevent the operation of a system that is necessary to reduce automobile emissions to the levels required by standards issued by the [Administrator] under section 202 of the act.
116 Cong.Rec. 32920 (1970) (emphasis added). Ethyl relies on the Senator’s subsequent repetition of the words “in and of themselves,” italicized above. Ethyl brief at 16. But the second paragraph makes clear that Senator Baker’s meaning is contrary to Ethyl’s implication. He is simply emphasizing the different nature of the two provisions and thus uses the words “in and of themselves” to show that under Section 211(c)(1)(A) the Administrator may act only because of the direct effects of lead additives on a legislative goal, protection of health, while under Section 211(c)(1)(B) he may act because of their indirect effects on another goal, implementation of emission control systems.61
Beyond these two points, Ethyl has nothing more to offer in support of its “in and of itself” reading. Thus it has made out no case at all, particularly in light of the realities of human exposure to lead and what Congress knew about those realities. Such consideration demonstrates both that, under Ethyl’s approach, EPA regulation of lead on health grounds would be impossible and that Congress could not possibly have intended the restrictive “by itself” reading. As has been discussed more extensively above, see pages 380-381 of 176 U.S.App.D.C., pages 8-9 of 541 F.2d supra, lead enters the human body from multiple sources, so that the effect of any one source is meaningful only in cumulative terms. If, for example, airborne lead were the only source of the lead body burden, and it caused, by itself, a blood lead level of 30 ug, there would be no danger to the public health. But if that hypothetical 30 ug is added to a possible 30 ug attributable to dietary ingestion, the blood lead level would be 60 ug, a definite threat to health. Under Ethyl’s approach, despite obvious endangerment such a cumulative finding is insufficient to justify regulation, Airborne, lead, in and of itself, may not be a threat. But the realities of human lead exposure show that no one source in and of itself (except possibly leaded paint) is a threat. Thus, under Ethyl’s tunnel-like reasoning, even if parallel legislation permitted regulation of other sources of lead exposure, which it does not, no regulation could ever be justified.
Such cannot be the case. Congress understood that the body lead burden is caused by multiple sources. It understood *403that determining the effect of lead automobile emissions, by themselves, on human health is of no more practical value than finding the incremental effect on health of the fifteenth sleeping pill swallowed by a would-be suicide.62 It did not mean for “endanger” to be measured only in incremental terms.63 This the Administrator also understood. He determined that absorption of lead automobile emissions, when added to all other human exposure to lead, raises the body lead burden to a level that will endanger health. He realized that lead automobile emissions were, far and away, the most readily reduced significant source of environmental lead. And he determined that the statute authorized him to reduce those emissions on such a finding. We find no error in the Administrator’s use of the cumulative impact approach.
4. Summary of the “Will Endanger” Determination. In sum, we must reject petitioners’ cramped and unrealistic interpretation of Section 211(c)(1)(A). Their reading would render the statute largely useless as a basis for health-related regulation of lead emissions. Petitioners’ arguments are rebuffed by the plain meaning of the statute and the Administrator’s interpretation of it,64 by the legislative history and the implications that can drawn from other sections of the same statute, by the relevant precedents, and by the established maxim that health-related legislation is liberally construed to achieve its purpose.65
We believe the Administrator may regulate lead additives under Section 211(c)(1)(A) when he determines, based on his assessment of the risks as developed by *404consideration of all the information available to him, and as guided by the policy judgment inherent in the statute, that lead automobile emissions significantly increase the total human exposure to lead so as to cause a significant risk of harm to the public health. Before so regulating, he must consider the possibility of regulation under Section 202.66 This interpretation of Section 211 does not allow for baseless or purposeless regulation, but does grant the Administrator the flexibility needed to confront realistically the public health problem presented by massive diffusion of lead emissions from automobiles.
B. Comparison with Substitute Additives
Even when the Administrator has determined that a fuel or fuel additive causes emissions which endanger the public health, he is not yet free to prohibit the substance under Section 211. He must first find, and publish his finding,
that in his judgment such prohibition will not cause the use of any other fuel or fuel additive which will produce emissions which will endanger the public health or welfare to the same or greater degree than the use of the fuel or fuel additive proposed to be prohibited.
Section 211(c)(2)(C), 42 U.S.C. § 1857f-6c(c)(2)(C).
Preliminarily it must be noted that the section requires a finding only before the Administrator “prohibits” a fuel or fuel additive under Section 211. Since the proposed regulations only “control” lead additives, the findings requirement, on its face, does not apply to the EPA action.67 If the requirement is read to apply, however, it plainly demands no more than the findings *405requirement of Section 211(c)(2)(B), as construed in Amoco, see page 395 of 176 U.S.App.D.C., page 23 of 541 F.2d supra. This conclusion is compelled by the identical genesis of the two provisions, plus the fact that the Section 211(c)(2)(C) finding is judgmental by its own terms. Thus where the judgment turns “on factual issues” we will “demand sufficient attention to these in the statement to allow the fundamental rationality * * * to be ascertained.” Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 180-181, 501 F.2d at 740-741. By contrast, where the judgment is necessarily more speculative, we will “demand adequate reasons and explanations, but not ‘findings’ of the sort familiar from the world of adjudication.” Id, 163 U.S.App.D.C. at 181, 501 F.2d at 741.68
After making the “will endanger” determination and the “substitute additives” finding, EPA has complied with the statutory mandate and is free to regulate the fuel or fuel additive under Section 211.
III. THE EVIDENCE
A. The Standard of Review
In promulgating the low-lead regulations under Section 211, EPA engaged in informal rule-making. As such, since the statute does not indicate otherwise, its procedures are conducted pursuant to Section 4 of the APA,69 5 U.S.C. § 553, *406and must be reviewed under Section 10 of the Act,70 5 U.S.C. § 706(2)(A)-(D). Our review of the evidence is governed by Section 10(e)(2)(A), which requires us to strike “agency action, findings, and conclusions” that we find to be “arbitrary, capricious, an "abuse of discretion, or otherwise not in accordance with law * * *.”71 5 U.S.C. § 706(2)(A). This standard of review is a highly deferential one. It presumes agency action to be valid. Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 415, 91 S.Ct. 814, 823, 28 L.Ed.2d 136, 152 (1971); Pacific States Box & Basket Co. v. White, 296 U.S. 176, 185-186, 56 S.Ct. 159, 163-164, 80 L.Ed. 138, 146-147 (1935); United States v. Chemical Foundation, 272 U.S. 1, 14-15, 47 S.Ct. 1, 6, 71 L.Ed. 131, 142-143 (1926).72 Moreover, it forbids the court’s substituting its judgment for that of the agency, Citizens to Preserve Overton Park v. Volpe, supra, 401 U.S. at 416, 91 S.Ct. at 823, 28 L.Ed.2d at 153, and requires affirmance if a rational basis exists for the agency’s decision.73 Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 290, 95 S.Ct. 438, 444, 42 L.Ed.2d 447, 458 (1974). Cf. United States v. Allegheny-Ludlum Steel Corp., 406 U.S. 742, 749, 92 S.Ct. 1941, 1946, 32 L.Ed.2d 453, 460 (1972).
This is not to say, however, that we must rubber-stamp the agency decision as correct. To do so would render the appellate process a superfluous (although time-consuming) ritual. Rather, the reviewing court must assure itself that the agency decision was “based on a consideration of the relevant factors * * *.”74 More*407over, it must engage in a “substantial inquiry” into the facts, one that is “searching and careful.” Citizens to Preserve Overton Park v. Volpe, supra, 401 U.S. at 415, 416, 91 S.Ct. at 823, 824, 28 L.Ed.2d at 152, 153. This is particularly true in highly technical cases such as this one. A court does not depart from its proper function when it undertakes a study of the record, hopefully perceptive, even as to the evidence on technical and specialized matters, for this enables the court to penetrate to the underlying decisions of the agency, to satisfy itself that the *408agency has exercised a reasoned discretion, with reasons that do not deviate from or ignore the ascertainable legislative intent.
Greater Boston Television Corp. v. FCC, 143 U.S.App.D.C. 383, 392, 444 F.2d 841, 850 (1970), cert. denied, 403 U.S. 923, 91 S.Ct. 2229, 2233, 29 L.Ed.2d 701 (1971). See also Essex Chemical Corp. v. Ruckelshaus, 158 U.S.App.D.C. 360, 367, 486 F.2d 427, 434 (1973), cert. denied, 416 U.S. 969, 94 S.Ct. 1991, 40 L.Ed.2d 558 (1974); Portland Cement Assn v. Ruckelshaus, 158 U.S.App.D.C. 308, 335, 486 F.2d 375, 402 (1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974); International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 444, 478 F.2d 615, 648 (1971).75
There is no inconsistency between the deferential standard of review and the requirement that the reviewing court involve itself in even the most complex evidentiary matters; rather, the two indicia of arbitrary and capricious review stand in careful balance. The close scrutiny of the evidence is intended to educate the court. It must understand enough about the problem confronting the agency to comprehend the meaning of the evidence relied upon and the evidence discarded; the questions addressed by the agency and those bypassed; the choices open to the agency and those made. The more technical the case, the more intensive must be the court’s effort to understand the evidence, for without an appropriate understanding of the case before it the court cannot properly perform its appellate function. But that function must be performed with conscientious awareness of its limited nature. The enforced education into the intricacies of the problem before the agency is not designed to enable the court to become a superageney that can supplant the agency’s expert decision-maker. To the contrary, the court must give due deference to the agency’s ability to rely on its own developed expertise. Market Street Railway v. Railroad Commission, 324 U.S. 548, 559-561, 65 S.Ct. 770, 776-777, 89 L.Ed. 1171, 1180-1182 (1945). The immersion in the evidence is designed solely to enable the court to determine whether the agency decision was rational and based on consideration of the relevant factors. Citizens to Preserve Overton Park v. Volpe, supra, 401 U.S. at 416, 91 S.Ct. at 823, 28 L.Ed.2d at 153; Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra, 419 U.S. at 285, 290, 95 S.Ct. at 441, 444, 42 L.Ed.2d at 455, 458. It is settled that we must affirm decisions with which we disagree so long as this test is met.76 Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra, 419 U.S. at 290, 95 S.Ct. at 444, 42 L.Ed.2d at 458; United States v. Allegheny-Ludlum Steel Corp., supra, 406 U.S. at 749, 92 S.Ct. at 1946, 32 L.Ed.2d at 460.
Thus, after our careful study of the record, we must take a step back from the agency decision. We must look at the decision not as the chemist, biologist or statistician that we are qualified neither by training nor experience to be, but as a reviewing court exercising our narrowly defined duty of holding agencies to certain minimal standards of rationality.77 “Although [our] in*409quiry into the facts is to be searching and careful, the ultimate standard of review is a narrow one.” Citizens to Preserve Overton Park v. Volpe, supra, 401 U.S. at 416, 91 S.Ct. at 824, 28 L.Ed.2d at 153. We must affirm unless the agency decision is arbitrary or capricious.78
With the “arbitrary and capricious” standard firmly in mind, we now turn to the evidence supporting the regulations before us.
B. Overview of the Evidence
Petitioners vigorously attack both the sufficiency and the validity of the many scientific studies relied upon by the Administrator, while advancing for consideration various studies allegedly supportive of their position. The record in this case is massive — over 10,000 pages. Not surprisingly, evidence may be isolated that supports virtually any inference one might care to draw. Thus we might well have sustained a determination by the Administrator not to regulate lead additives on health grounds. That does not mean, however, that we cannot sustain his determination to so regulate. As we have indicated above, we need not decide whether his decision is supported by the preponderance of the evidence, nor, for that matter, whether it is supported by substantial evidence.79 To the contrary, we must sustain if it has a rational basis in the evidence. Keeping in mind the precautionary “will endanger” standard under which the Administrator acted, we have no difficulty in terming his decision rational.
A word about our approach to the evidence may be in order. Contrary to the apparent suggestion of some of the petitioners, we need not seek a single dispositive study that fully supports the Administrator’s determination. Science does not work that way; nor, for that matter, does adjudicatory fact-finding. Rather, the Administrator’s decision may be fully supportable if it is based, as it is, on the inconclu*410sive but suggestive results of numerous studies. By its nature, scientific evidence is cumulative: the more supporting, albeit inconclusive, evidence available, the more likely the accuracy of the conclusion.80 If, as petitioners suggest, one single study or bit of evidence were sufficient independently to mandate a conclusion, there would, of course, be no need for any other studies. Only rarely, however, is such limited study sufficient. Thus, after considering the inferences that can be drawn from the studies supporting the Administrator, and those opposing him, we must decide whether the cumulative effect of all this evidence, and not the effect of any single bit of it, presents a rational basis for the low-lead regulations.
While we have studied the record with great care, we do not discuss it all here; to do so would make this already lengthy opinion completely unwieldy. Instead, we shall briefly review the bases for the Administrator’s conclusions that petitioners have singled out for special attack.81 Before we turn to those issues, however, we should note that some things appear to be uncontested. Thus petitioners seem to concede the following: that lead serves no known purpose in the human body; that lead in sufficiently high quantity is destructive to the body, causing anemia, severe intestinal cramps, paralysis, neurologic damage, and, in sufficient dosage, death, Third Health Document at III — 1, 2, JA 54-55; that more than 250,000 tons of lead per year are used in production of lead additives, accounting, according to EPA, for approximately 90 percent of all airborne lead,82 id. at Table II — 1, JA 46; that lead concentrations in the air over our largest cities are 2,000 times greater than lead concentrations in the air over the mid-Pacific, NAS Report at 205; that lead in the ambient air contributes to body blood lead levels, Nalco Supp.Br. at 37; supplemental brief of petitioners PPG Industries and E. I. duPont de Nemours & Company (hereinafter PPG/duPont Supp.Br.) at 23; and that blood lead levels are a reasonable indication of the body’s lead burden. Stripped of their generalized and largely unsubstantiated claims of “bias” and “distortion of the evidence,” petitioners principally challenge three EPA conclusions: (1) that, based on a preliminary determination that blood lead levels of 40 ug are indicative of danger to health, elevated blood lead levels “exist to a small but significant extent in the general adult population, and to a very great extent among children,” Third Health Document at VII-3, JA at 144; (2) that airborne lead is directly absorbed in the body through respiration to a degree that constitutes a significant risk to public health; and (3) that airborne lead falls to the ground where it mixes with dust and poses a significant risk to the health of urban children.
1. Blood Lead Levels are Elevated Among the General Public
a. Blood lead levels of 40 ug are indicative of danger to health. Although recognizing that a blood lead level of 40 ug “does not represent a sharp demarcation between health and disease,” the Administrator found it “prudent to regard blood lead levels over 40 ug/100g as indicators of lead intake that should be prevented.” Third Health Document at III — 11, JA 64. Petitioners contest this determination. Much of their argument, however, adds up to nothing more than that they are unhappy the Administrator chose to exercise his judgment and think the statute does not permit such discretion. Petitioners cite medical studies that show not that a blood lead level *411of 40 ug is unrelated to danger, but that scientists are simply uncertain about the effect of such blood levels. Ethyl brief at 37-38 and Supp.Br. at 41-42; PPG/duPont brief at 11-12; Nalco brief at 14. But, as we have seen, exercise of reasoned discretion based on the evidence is not only permitted, but mandated, by Section 211(c)(1)(A). The administrator recognized from the outset that the health effects of varying human blood lead levels were uncertain and that continuing research was necessary.83 His determination of the 40 ug level is, on its face, an assessment of risks based on the known facts and not improper as such.
In addition to their generalized attack on the Administrator’s power, petitioners challenge the sufficiency of the evidence to support his determination. Under the “will endanger” standard, however, we find the evidentiary basis for the Administrator’s determination to be more than adequate. Petitioners do not contest the recommendation of the United States Public Health Service that 80 ug be taken as the standard of unequivocal lead poisoning, or the Service’s recommendation that blood lead levels of 50-79 ug justify immediate evaluation for possible lead poisoning. Medical Aspects of Childhood Lead Poisoning, HSMHA Health Reports, 86(2), 140-143 (1971), cited in Third Health Document at IV-3, JA 71. What draws petitioners’ fire is only the last of the Service’s recommendations, adopted by EPA, that for older children and adults “a blood lead concentration of 40 ug or more per 100 ml of whole blood * * * be considered evidence suggestive of undue absorption of lead, either past or present.” Id. While the reasonableness of establishing as a danger point a lead exposure level somewhat lower than that at which actual damage may occur seems both logically unassailable and well within the intent of the “will endanger” standard, the Administrator went beyond that sufficient conclusion. He recognized, and warned of, possible dangers, particularly to children, from lead at the 40-60 ug level, and even below, thereby suggesting that the Public Health Service’s recommended danger level, designed for adults, may be set too high. Because this latter evidence was only suggestive, however, the Administrator conservatively settled on the 40 ug standard.84 Third Health Document III — 5, 6, 7, 8, JA 58-61. Such a “prudent” determination is well within the Administrator’s discretion under the “will endanger” standard. We have examined the evidence relied upon carefully and, while it is not necessary to summarize it here,85 we *412find that it provides a rational basis for the Administrator’s determination.
b. Blood lead levels are elevated among a small, but significant, number of adults and a considerable number of children. Again, petitioners challenge the Administrator’s determination as unsupported by the evidence. The problem here is one of choosing among the items of evidence. Petitioners rely heavily on the results of the so-called Seven Cities Study,86 which found a very small percentage of adults with elevated (in excess of 40 ug) blood levels.87 PPG/duPont brief at 12-15; Nalco brief at 16. The Administrator, on the other hand, finds serious methodological flaws in the Seven Cities Study that limit its usefulness, 38 Fed.Reg. 33735, and relies instead on studies which concededly support his conclusion, Third Health Document at Tables VII — 1, 2 & 3, JA 145-147, but which petitioners score as representative only of certain occupational groups. PPG/duPont brief at 15; Nalco brief at 15-16.
Having analyzed this evidence and the arguments of the parties, we would again defer to the Administrator’s judgment. First we note that, while contesting the source of lead exposure, petitioners do not challenge at all the Agency’s conclusion that blood lead levels are elevated in a large number of children, including a possible 25 percent of' all preschool children living in substandard housing.88 Third Health Document at Table VII-3, JA 147.
Next, while the studies relied on by the Administrator are largely of various occupational groups, they are frequently occupations whose only exposure to lead is through the ambient air in which their workers — policemen, mailmen, service station employees, parking lot attendants, and the like — are forced to spend their working hours.89 Third Health Document at Table *413VII-1, JA 145. Contrary to the arguments of petitioners, studies of occupational groups are often particularly valuable in acting as an early warning system of possible effects on the public at large. The intensive exposure in particular occupations essentially accelerates the effects of long-term exposure of the general public and provides an identifiable, highly exposed, and readily accessible group of research subjects. See D. Clark & B. MacMahon, Preventive Medicine 26 (1967) (hereinafter Clark & MacMahon). Thus much of the evidence in Reserve Mining of the danger to the general public of absorption of asbestos particles came from epidemiological studies of asbestos workers. Reserve Mining Co. v. EPA, supra, 514 F.2d at 507.90 Because of this the occupational data are not only in themselves indicative of high blood lead levels among a significant portion of the adult population, but they suggest the future likelihood, absent controls, of increasing high blood lead levels among even more widely dispersed portions of the public. In any case, however, the evidence relied upon by the Administrator was not only of occupational groups. Some evidence of high blood lead level relates to such “specialized” groups as “Drivers of Cars,” “Male Commuters,” and “Urban” residents. Third Health Document at Tables VII-1 & 2, JA 145-146.
Petitioners argue that the negative conclusions of the Seven Cities Study and several other studies should outweigh the posifive indications of the studies described above. PPG/duPont Supp.Br. at 16-23; NPRA Supp.Br. at 5-8. The Administrator disagreed, and we cannot fault his conclusion. Because of the many uncontrolled variables in epidemiological studies, valid conclusions may be obscured. But, while the possibility of false negative conclusions is a real one, the possibility of false positive findings is considerably less likely. Clark & MacMahon, supra, at 100. Thus the Administrator credited the occupational and other studies reporting a small, but significant, portion of the general population with elevated blood lead levels and discounted the negative studies on which petitioners rely. Third Health Document at VII-3, JA 144. We cannot say his conclusion was arbitrary or capricious.
2. Automobile Lead Emission Products are Directly Absorbed in the Body to a Significant Extent
Since it is apparent from the face of his decision and the Third Health Document that the Administrator considered all the evidence before him, the only issue is whether he treated that evidence in a rational manner. Petitioners have now conceded that lead emissions are directly absorbed in the body from the ambient air, and they challenge only whether the extent of absorption is significant enough to justify these regulations. Nalco Supp.Br. at 37; PPG/duPont Supp.Br. at 23. The Adminis*414trator’s conclusion that lead absorption from the air is significant is amply supported by the record in this rule-making.
The Administrator relied on three types of evidence: theoretical, epidemiological, and clinical studies. The theoretical evidence consisted of a set of calculations designed to estimate the amount of lead in the air which, when added to an average dietary intake, would suffice to bring the blood lead burden of a “standard man” up to 40 ug.91 These calculations indicated that the 40 ug level can be reached by exposure to ambient air lead concentrations no greater than those now found in parts of our larger cities. Significantly, the results of two clinical experiments support the estimates derived from these theoretical calculations. The calculations, as well as petitioners’ misguided attacks on their validity, are discussed in Appendix A to this opinion. See pages 427 to 428 of 176 U.S.App.D.C., pages 55 to 56 of 541 F.2d infra.
The second type of evidence relied upon by the Administrator in reaching his conclusion that airborne lead contributes significantly to the human lead body burden consisted of epidemiological research. Epidemiologists study the effects of various phenomena on humans under uncontrolled or “natural” conditions. These effects are correlated with other observed facts in' an attempt to develop significant relationships among the data. The science is limited, however, in that inferences may be drawn, but relationships cannot be proved by the correlations alone. JA 583. The studies before the Administrator were of large groups of people; correlations were sought between blood lead level and exposure to lead in the ambient air. The studies were confounded, however, by the multiple sources of lead. Since diet accounts for a major portion of the body lead burden, an individual’s blood lead level varies not only according to his exposure to lead in the ambient air, but according to his daily dietary intake of lead. Wide variations in dietary lead intake, which are common, can completely mask the effects of air lead absorption. Nonetheless, none of the epidemiological studies could control or measure dietary lead intake. This uncertainty in the data severely limited the usefulness of the broadly conceived epidemiological studies and led the Administrator to rely instead on data limited to situations in which dietary exposure could roughly be termed constant.
Following this rationale, the Administrator focused on the consistent relationship found between air and blood lead levels within particular metropolitan areas, rather than on the lack of such a relationship between areas.92 See pages 429 to 430 of *415176 U.S.App.D.C., pages 57 to 58 of 541 F.2d infra. The Administrator also drew support for his conclusion that lead in the air significantly affects lead in the blood from studies conducted in single neighborhoods. There, too, confounding factors were minimized by proximity, and there, too, a clear direct relationship was found. See pages 431 to 433 of 176 U.S.App.D.C., pages 59 to 61 of 541 F.2d infra. Thus the epidemiological studies, although perhaps insufficient to justify the Administrator’s decision if considered singly or even collectively, were reasonably relied on as part of the basis for the low-lead rules.
The conclusions the Administrator drew from theoretical calculations and the epidemiological studies are significantly bolstered by two important clinical studies. Clínica, studies are laboratory experiments on humans in which variables can be controlled and causal relationships demonstrated. Clinicians attempt to reproduce atmospheric conditions in the laboratory and measure the response to pollution of small numbers of humans chosen to represent the population at large. JA 583. Because of its necessarily significant intrusion on the subjects’ lives, clinical research is limited in the number of subjects it can study. Nonetheless, the two clinical studies are especially valuable since they are the only attempts to control or account for dietary lead intake. Thus they sought to measure the contribution of respiratory absorption of lead unmasked by variations in dietary lead. Both studies found that airborne lead provided a significant portion of the lead in the blood of the experiments’ subjects. Moreover, the amount of lead actually absorbed from the air during these studies corresponded closely with the amount EPA’s theoretical calculations predicted would be absorbed. See pages 433 to 435 of 176 U.S.App.D.C., pages 61 to 63 of 541 F.2d‘ infra.
While we must consider petitioners’ arguments carefully, the necessary existence of limitations and inconsistencies in the data remind us to observe carefully our limited appellate function. It does not matter whether or not we agree with the Administrator’s determination. Nor does it matter whether the evidence might support a conclusion contrary to that reached by the Administrator. All that is of concern to us is that there be a rational basis in the evidence for the conclusion reached. We cannot say there is not. The bulk of the evidence cited supports the Administrator. See Appendix A to this opinion. The Administrator treated all the evidence in a consistent and rational manner. This treatment disposed of most of the studies relied upon by petitioners, leaving the plain conclusion that lead emissions form a significant part of the human body burden. Particularly in light of the precautionary nature of the “will endanger” standard, we cannot find the Administrator’s conclusion to be arbitrary or capricious. Accordingly, we must uphold his determination.
3. Lead Exposure from Dustfall Threatens the Health of Children
While we would have no difficulty in sustaining the low-lead regulations solely on the basis of the evidence and conclusions discussed above, the Administrator based his decision to regulate on other evidence as well. He presented a hypothesis, which he found consistent with known information, that urban children are particularly threatened by lead additives in that they are prone to ingest lead emissions that have fallen to the ground and mixed with dust. While the hypothesis is admittedly not proved as fact, we need not decide whether it would be sufficient by itself to support the low-lead regulations, for it is offered only in support of the evidence already presented. Petitioners vigorously attack the hypothesis nonetheless, but constantly confuse the issue by ignoring the flexibility *416of the “will endanger” standard and by attempting to refute the Administrator with evidence that he both understands and has accounted for. Thus it is helpful first to set these matters straight.
First, as we have demonstrated above, see pages 384-392 of 176 U.S.App.D.C., pages 12-20 of 541 F.2d supra, the “will endanger” standard is a precautionary standard that embraces a wide range of permissible proof. It is therefore no objection to the dustfall hypothesis that it is merely a hypothesis. A supportable and reasonable hypothesis may well form the basis for regulations under Section 211(c)(1)(A). Indeed, the totality of evidence relied upon in the Reserve Mining case constituted no more than such a hypothesis. See pages 417-418 of 176 U.S.App.D.C., pages 45-46 of 541 F.2d infra.
Second, although it seems too obvious to mention, the Administrator understood and expressly recognized that leaded paint “is the primary cause of clinical lead poisoning” in children, 38 Fed.Reg. 33735. Although petitioners repeatedly try to rebut EPA conclusions by stating this fact, the Administrator was well aware of it and his evaluation of the data before him invariably accounted for it. We do not believe the importance of lead paint to the lead poisoning problem in children invalidates the Administrator’s reliance on the dustfall theory. The Administrator justified his hypothesis as a basis for regulátion on grounds that dustfall may be regulated far more readily than leaded paint, that dust-fall contributes significantly to the threat from leaded paint by raising blood lead levels so as to make lead poisoning easier to contract, and that even when acute lead poisoning does not develop absorption of lead dustfall may cause undesirable and avoidable subclinical effects. 38 Fed.Reg. 33736-33737.93 We have held, see pages 401-403 of 176 U.S.App.D.C., pages 29-31 of 541 F.2d supra, that concern about the cumulative impact of exposure to various lead sources may justify regulation of lead additives under Section 211(c)(1)(A). Thus we cannot say that these reasons for regulation, when added to the demonstrated danger to the general public from respiratory absorption of lead emissions, see Appendix A to this opinion, and if supported by the evidence, do not form a permissible basis for the low-lead regulations under the “will endanger” standard.
With these preliminaries settled, we turn to the evidence supporting the Administrator’s determination. Since the hypothesis is only a secondary basis for the regulations, and since the evidence cited goes considerably beyond that necessary to establish the hypothesis as reasonable, we shall address only the thrust of petitioners’ objections and the Administrator’s responses. The logical steps to the Administrator’s conclusion are these:
a. High lead concentrations in dust and dirt are prevalent in urban areas.
b. In most circumstances, lead from exhausts and not lead paint or lead from stationary sources is the primary source of lead in urban dust and dirt.
c. Children prone to pica, about 50 percent of those between the ages of one and three, eat nonfood objects, including dust and dirt.
d. As a result of ingesting dust and dirt contaminated with lead fallout, children can be expected to absorb lead into their bodies.
38 Fed.Reg. 33736. If the intermediate steps are supported by the evidence, the validity of the Administrator’s conclusion as a reasonable hypothesis is unassailable. Our study of the underlying evidence convinces us that it is firm and convincing, and certainly sufficient to support the Administrator’s hypothesis as reasonable.
Petitioners concede that lead concentrations are high in the dust near highways and adjacent to homes with lead paint. They contest, however, that high lead con*417centrations are otherwise prevalent in urban areas. PPG/duPont Supp.Br. at 37 n. 86. The facts rebut their argument. The NAS Panel found that “[t]he concentration of lead in street dust and surface soil of large cities is extremely high.” NAS Report at 30 (emphasis added). This conclusion was based on a finding that, compared to the usual range of dust lead concentrations of 2-200 parts per million (ppm), dust lead concentrations in cities averaged 1,636 ppm and 2,413 ppm respectively for residential and commercial sites. Even in city parks lead concentrations ranged from 194 ppm to 3,357 ppm. Id. Other evidence of record shows high lead concentrations in Central Park Zoo in New York City, JA 2630, and school playgrounds in Philadelphia, JA 674.
Where does this high dust lead content come from? The Administrator concluded that since 90 percent of the lead in the ambient air is from automotive exhausts and since, as petitioners concede, PPG/du-Pont Reply Br. at 17, the lead eventually settles to the ground, most of the lead in dust is a product of automobile lead emissions. Petitioners do not seriously contest this; indeed their own studies suggest this by showing that lead content in soil decreases with distance from highways.94 Rather, they argue that “lead from paint or stationary sources is the source of lead in dust and dirt in areas where young children play.” PPG/duPont Supp.Br. at 37 n. 86. This argument reflects petitioners’ consistent refusal to recognize that city children play regularly in city streets, where petitioners concede lead dustfall from automobiles accumulates. PPG/duPont Reply Br. at 17. Since the Administrator’s concern about dustfall is based on the harm it may cause city children, petitioners’ argument that children are not exposed to automobile dustfall where they play must be rejected.
Petitioners do not contest that pica is a common phenomenon among preschool children, although they note that it “is a psychological disorder; it is not characteristic of all children * * PPG/duPont Supp.Br. at 37 n. 86. While this is certainly true, pica is characteristic of a significant number of children. Petitioners do not challenge the conclusion of the NAS Panel that pica occurs in at least 50 percent of both middle- and lower-class children. NAS Report at 133.
On the basis of this evidence alone it is reasonable to hypothesize that children with pica will ingest dust and dirt containing lead dustfall from automobiles. Indeed, on such limited evidence the NAS Panel accepted the dustfall hypothesis as credible:
Airborne lead wastes from such sources as automobile emissions and the weathering and demolition of old buildings can be expected to have a significant additive effect on the total intake. This would be sufficient to evoke compensatory metabolic responses that are now considered subclinical (such as increased urinary ALA), at the very least. It may be estimated that dustfall from airborne lead, if swallowed, can make a significant contribution to a small child’s total lead intake and thereby contribute to the occurrence of lead poisoning, especially in urban areas. Even so, the direct ingestion of lead-pigment paints is clearly the principal environmental source in cases of severe acute lead poisoning in young children.
NAS Report at 140 (emphasis added). Again petitioners focus on the last sentence and emphasize the danger from leaded paints. Since the Administrator is in agreement with them on the point, however, their arguments are superfluous.
Petitioners’ primary claims are not addressed to the evidence that establishes the dustfall hypothesis as tenable. Rather, they argue that there is no evidence that lead dustfall from automobiles is in fact swallowed by children with pica. While such proof is not necessary to establish the Administrator’s hypothesis as reasonable, see pages 416-417 of 176 U.S.App.D.C., *418pages 44-45 of 541 F.2d supra, we note that a considerable amount of circumstantial evidence does support his conclusion. See Appendix B to this opinion.
In any case, all the evidence suggesting that the children with pica tend to eat dust contaminated with lead fallout from automobiles takes the Administrator’s theory far beyond the hypothesis stage. He offered only a tentative result, “a hypothesis consistent with information provided by a variety of studies.” 38 Fed.Reg. 33736. Undoubtedly, he has shown that. Indeed, as reasonable medical hypotheses go, this one is particularly solid. This is vividly demonstrated by recalling the kind of evidence relied upon to justify similar precautionary relief in Reserve Mining.
The question in that case was the validity of the hypothesis that ingestion of asbestos fibers was dangerous to health. The reason for concern was that epidemiological studies had associated inhalation of asbestos with cancer. The evidence supporting the hypothesis was of three kinds: (1) a court-sponsored study to determine whether asbestos fibers were present in residents who drank the polluted water; (2) animal studies designed to measure whether asbestos fibers can be absorbed into the body from the stomach; and (3) epidemiological studies associating inhalation of asbestos fibers with gastrointestinal cancer and the theory that this may be due to ingestion of asbestos fibers initially inhaled. Reserve Mining Co. v. EPA, supra, 514 F.2d at 514. Upon review of the evidence the court concluded (1) asbestos fibers were not present in longtime residents; (2) the animal studies were ambiguous on whether asbestos fibers could be absorbed if ingested; and (3) the theory that asbestos workers ingested asbestos fibers as well as inhaled them was no more than a theory, id. at 514-516. Nonetheless, solely on the basis of this evidence, the court concluded that “the theory that excess cancers may be attributed to the ingestion of asbestos fibers rests on a tenable medical hypothesis,” id. at 516. On the basis of this tenable hypothesis, the court accepted the further hypothesis that asbestos could be ingested from the drinking water and that, therefore, there was a “reasonable medical concern” for the public health which justified abatement of the asbestos discharge under the “endangering” language of the FWPCA. Id. at 520.
None of these uncertainties cloud the Administrator’s dustfall hypothesis. Lead is present in children, and in elevated amounts. Children do ingest dust, and dust is heavily laden with lead. Animal studies prove that ingested lead dust is absorbed into the bloodstream. And epidemiological studies associate high lead concentrations in children with high lead concentrations in dirt and dust, and with proximity to automobiles. Since the automobile is the predominant source of lead in dust, the Administrator’s hypothesis stands firm as reasonable, undoubtedly with more support in studies already made than the hypothesis that justified regulation in Reserve Mining. Indeed, the primary difference between this case and Reserve Mining is that the Eighth Circuit justified ordering abatement of as-, bestos discharges into the water solely on the basis of a hypothesis,95 while here the hypothesis is offered only as support for the regulations, the primary basis being the demonstrated danger to health posed by inhalation of lead emissions. In this context, and keeping in mind the precautionary nature of the “will endanger” standard, we have no difficulty in finding the dustfall hypothesis sufficiently well grounded to support the Administrator’s limited reliance on it.96
*419C. Summary of the Evidence
From a vast mass of evidence the Administrator has concluded that the emission products of lead additives will endanger the public health. He has handled an extraordinarily complicated problem with great care and candor. The evidence did not necessarily always point in one direction and frequently, until EPA authorized research, there was no evidence at all.97 The Administrator reached his conclusion only after hearings spread over several months, consideration of thousands of pages of documents, publication of three health documents, three formal comment periods, and receipt of hundreds of comments. Each study was considered independently; its worth was assessed only after it was measured against any critical comments. From the totality of the evidence the Administrator concluded that regulation under Section 211(c)(1)(A) was warranted.98
In tracking his path through the evidence we, in our appellate role, have also considered separately each study and the objections petitioners make thereto. In no case have we found the Administrator’s use of the evidence to be arbitrary or capricious. Having rejected the individual objections, we also reject the overall claim of error. We find the Administrator’s analysis of the evidence and assessment of the risks to be well within the flexibility allowed by the “will endanger” standard. Accordingly, we affirm his determination that lead emissions “present a significant risk of harm to the health of urban populations, particular*420ly to the health of city children.” 38 Fed. Reg. 33734.
IV. EPA PROCEDURES
Petitioners’ last major objection to the low-lead regulations is the claim that they are procedurally defective in that, after allowing a total of three formal comment periods, the Administrator did not allow a fourth prior to issuing the regulations.99 Petitioners claim this deprived them of administrative due process in contravention of the demands of Section 4 of the Administrative Procedure Act, 5 U.S.C. § 553 (1970). Ethyl Br. at 50; PPG/duPont Br. at 43; Nalco Br. at 64; NPRA Br. at 44. We find the argument to be without merit.
Section 4 requires that prior to final promulgation of a rule the agency must make public “either the terms or substance of the proposed rule or a description of the subjects and issues involved.” 5 U.S.C. § 553(b)(3). The courts have added useful flesh to this statutory language. The notice should be sufficiently descriptive of the “subjects and issues involved” so that interested parties may offer informed criticism and comments. See, e. g., Portland Cement Assn v. Ruckelshaus, 158 U.S.App.D.C. 308, 325-327, 486 F.2d 375, 392-394 (1973), cert denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974); Mobil Oil Corp. v. FPC, 157 U.S.App.D.C. 235, 248 n.39, 483 F.2d 1238, 1251 n.39 (1973). But the notice need not contain “every precise proposal which [the agency] may ultimately adopt- as a rule.” California Citizens Band Assn v. United States, 375 F.2d 43, 48 (9th Cir.), cert. denied, 389 U.S. 844, 88 S.Ct. 96, 19 L.Ed.2d 112 (1967). This last qualification is important since the notice invites comments and the comments will frequently prompt changes in the ultimate regulations.
There is nothing in Section 4 that requires new notice whenever the agency responsibly adopts the suggestions of interested parties. Nonetheless, in this case the Administrator did just that. When, after two comment rounds, criticism of the regulations originally proposed prompted him to alter somewhat the theories on which he was acting, he did not simply promulgate final rules based on these new theories. Rather, he reproposed the regulations as amended and opened them up for a third comment round. It was only when the final regulations were issued that the Administrator omitted a fourth' comment period. There was no reason for further comment. The Agency theories underlying the final regulations were identical with those of the Second Health Document and reproposed regulations. The Third Health Document differs from the Second only in that it incorporates new information received since the reproposal and responds to the comments. The only change in the regulations as issued was the switch, supported by comments from a majority of refiners, from *421leaded pool averaging to total pool averagingr100
All significant new information developed during the rule-making in this area on the frontiers of scientific knowledge was made available to petitioners and the public for comment well in advance of issuance of the final regulations on November 28, 1973.101 Thus both the requirements and the spirit of Section 4 were complied with. Nevertheless, the dissent vigorously and at length102 attacks the Administrator’s use of several studies, alleging they were not made available for comment. A brief review of the record relating to these studies demonstrates that this attack is without merit.
1. The Pilot Lead Isotope Study. The pilot lead isotope study consists of two separately reported experiments which produced similar results.103 One of these experiments was presented in a paper delivered at the October 1-2, 1973 Conference on Low Level Lead Toxicity sponsored by EPA and HEW and has apparently not been published elsewhere. However, the results of this experiment were reported to EPA in a letter dated August 28,1972, and placed in the public file at that time. Doc. 875.104 More significantly, a draft report of the compan*422ion experiment was received by EPA and placed in the public file in early May of 1973. Doc. 470, dated May 5, 1973. The final version of this study, which was eventually published in November 1973, was made available to Ethyl and placed in the public file in early August 1973.105 Thus the pilot lead isotope study was available for comment and criticism well before the regulations were promulgated.106
2.The Unpublished Japanese Study. The dissent places great stress on the fact that this unpublished and undated draft107 bears an inscription prohibiting citation or quotation. Dissent at 452 of 176 U.S.App.D.C., at 80 of 541 F.2d. We note first that since the study was prepared and submitted to EPA by the International Lead Zinc Research Organization, an industry group including these petitioners, Government Supp.Br. at 46, their claim of prejudice from the Administrator’s use of the study rings false. Moreover, the study was placed in the public file and sent to petitioner Ethyl some four months before the Administrator’s decision was announced.108
3. The Chamber Study. The dissent attempts to build an argument from the fact that the Second Health Document cites the chamber study109 as a “preprint” whereas the Third references the study to the printed proceedings of the international symposium at which the paper was delivered. Dissent at 452, 455 of 176 U.S.App.D.C., at 80, 83 of 541 F.2d. ' But see id. at 452-453 of 176 U.S.App.D.C., at 80-81 of 541 F.2d. However, comparison of the “preprint” with the final publication shows that they are absolutely identical.110 Thus the chamber study was in the public realm from the beginning of the formal comment period on the reproposed regulations.
4. The Seven Cities Study. The dissent criticizes only the Administrator’s reliance on an EPA reanalysis of the data on which this study was based, since it is conceded that the study itself111 was available at the beginning of the formal comment period. Dissent at 452-453, 455 of 176 U.S.App.D.C., at 80-81, 83 of 541 F.2d. Three facts vitiate this criticism of the Administrator’s reliance on the Sev*423en Cities Study. First, the dissent is simply incorrect when it implies that the Administrator relied only on the “additional analysis.” Dissent at 455 of 176 U.S.App.D.C., at 83 of 541 F.2d. The Administrator actually relied primarily on a finding reported in the original study. See page 414 of 176 U.S.App.D.C., & note 92 supra, page 42 of 541 F.2d, & note 92 supra. The reanalysis merely reconfirmed the validity of that finding.112 Second, the reanalysis itself is just EPA’s effort to reexamine the study in light of the critical comments received on it.113 It would be unreasonable to require another formal comment period whenever an agency determines that comments received during a preceding comment period do not undermine the validity of a particular piece of evidence. Finally, EPA made the reanalysis public in February 1973.114
5. The Newark, Rochester, Philadelphia, and Chicago Studies. The only clear references to these studies,115 which bear on the *424validity of the dustfall hypothesis, occur in a portion of the Administrator’s decision that addresses the question, “What new information has become available since reproposal of the regulation and as a result of the additional comment period?” JA 5. This section of the decision follows the Administrator’s explanation and justification of his conclusion that “lead particle emissions from motor vehicles present a significant risk of harm to the health of urban populations, particularly to the health of city children.” JA 2. See JA 2-5. The Administrator was required to take account of the new information by the statute which directs him to consider “all relevant medical and scientific evidence available to him.” Section 211(c)(2)(A), 42 U.S.C. § 1857f-6c(c)(2)(A) (emphasis added).
Moreover, as the Administrator’s placement of the discussion of these studies makes clear, the studies are not needed or used to support the conclusion that the dustfall hypothesis is reasonable. That conclusion is a sufficient basis for regulation under the statute. See pages 383-392, 416 of 176 U.S.App.D.C., pages 11-20, 44 of 541 F.2d supra. These additional studies, which are corroborative of the dustfall hypothesis, see pages 416-118 of 176 U.S.App.D.C., pages 44r-46 of 541 F.2d supra and Appendix B to this opinion, play no role in the Administrator’s decision to regulate.116
Finally, petitioners and the dissent suggest that even if all the new studies were made available to the public for comment well before the Administrator reached his decision, Section 4 of the APA requires still another procedural step: that the Administrator, before arriving at his decision, publicly identify the recently received studies and comments on which he intended to rely so that the public might have yet another opportunity to comment on that material. If this new round of comments provided any significant information, the Administrator would presumably be able to use that information only if he formally stated his intent to do so and allowed still another round of comments. For obvious reasons, there is no support for this novel suggestion in the Act or in the jurisprudence.117
*425The record in this case clearly demonstrates that EPA fully satisfied the requirements of administrative due process. In fact, EPA’s efforts to elicit informed comment on its proposed action went far beyond the measures it was required to take. All health-related documents, including internal EPA policy memoranda, were made public upon receipt,118 and comments on the documents were accepted until the date of final promulgation.119 These documents included drafts of the Administrator’s decision whose contents were very similar to the version finally published120 and a draft of the Third Health Document which was substantially identical to the final draft.121 Both the draft of the Administrator’s decision and the regulations and the draft of the Third Health Document were circulated for comment, and comments were received122 and acted upon.123 On this record, we cannot find that petitioners were deprived of administrative due process by EPA procedures.124
*426V. CONCLUSION
The complex scientific questions presented by this rule-making proceeding were “resolved in the crucible of debate through the clash of informed but opposing scientific and technological viewpoints.” International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 448, 478 F.2d 615, 652 (1973) (concurring opinion of Chief Judge Bazelon). On January 31, 1971 the EPA began the debate by publishing advance notice of proposed rule-making concerning possible controls on lead additives in gasolines because of their possible danger to health. On February 23, 1972 it published the proposed regulations supported by a document, Health Hazards of Lead (First Health Document), prepared by the EPA scientific staff. It invited comment from the lead industry, the scientific community, and the concerned public. The EPA held public hearings in Washington, D.C., Dallas, and Los Angeles to give people across the country an opportunity to join the debate.
On January 10, 1973 the EPA reproposed the regulations in slightly changed form, supported by a Second Health Document which reflected the scientific comments on the first and brought the scientific information on the subject up to date. Scientific studies, pro and con, which had become available since the proposed regulations were first published were included. Again the EPA invited the parties, the scientific community, and the concerned public to comment. Finally, on November 28, 1973, *427almost three years after the debate was joined, the EPA promulgated its regulations accompanied by a 10,000-word opinion, thoroughly and comprehensively analyzing the various scientific studies and giving its reasons why it resolved the scientific debate it had provoked in favor of protecting the public from the danger of lead emissions. A Third Health Document, extensively detailing and reviewing the current state of scientific knowledge of the health effects of airborne lead, also accompanied the regulations and the reasons for their issuance.
Because of the importance of the issues raised, we have accorded this case the most careful and exhaustive consideration. We find that in this rule-making proceeding the EPA has complied with all the statutory procedural requirements and that its reasons as stated in its opinion provide a rational basis for its action. Since we-reject all of petitioners’ claims of error the Agency may enforce its low-lead regulations.
Affirmed.
APPENDIX A
Evidence Relating to the Administrator’s Conclusion That Automobile Lead Emission Products Are Directly Absorbed in the Body to a Significant Extent.
1. Experimental Calculations. The Administrator leads off his discussion of the relationship between airborne and blood lead levels with a set of calculations based on the experimental work of Kehoe and designed to provide a rough estimate of the dangers posed by exposure to various ambient air lead concentrations. Third Health Document, V — 2 to V — 9, JA 82-89. Kehoe, in his classic Harben Lectures,1 reported on his work in supplementing the daily lead intake of various subjects by differing amounts of lead and carefully measuring their blood lead levels over time. Subject S.W.’s normal dietary intake of 300 ug lead was supplemented by an additional 300 ug, and this total daily dietary intake of 600 ug brought him to a blood lead level of 40 ug. Since the 40 ug mark is EPA’s warning level, the Agency sought to calculate, based on Kehoe’s data, the amount of airborne lead needed to be absorbed to reach this level.
These calculations start with the absorption figures for ingested lead. Commonly accepted data for daily dietary intake for adults are on the order of 200-300 ug/day, with ranges of 100 to 500 ug/day. Of this amount, 10 percent is thought to be absorbed into the blood; the remainder is excreted. Thus the 600 ug daily intake of subject S.W. caused a daily absorption of 60 ug into the blood. If daily dietary lead intake averages 200-300 ug, for a daily absorption of 20-30 ug, what air lead exposure level, asked EPA, could result in an increased absorption of 30-40 ug daily, so that total absorption remained 60 ug? Third Health Document at V-6, JA 86.
Absorption of respiratory lead is dependent on three factors: air lead concentration, volume of air inhaled daily, and percent of inhaled lead absorbed. EPA based its calculations on daily respiration of 15 and 20 cubic meters (m3), and absorption of 17, 30, and 37 percent of the lead in the respired air. By this process, the Agency calculated that a “standard” man, who breathes 20 m3 of air daily, and absorbs 30 percent of the lead in the respired air would absorb 30 ug of lead daily if exposed to air lead concentrations of 5.0 ug/m3, a concentration found in parts of several of our larger cities. At this rate of respiratory absorption, a subject absorbing 30 ug of dietary lead daily would reach a blood lead level of 40 ug within a year. Third Health Document at V — 4 to V-7, Figures V-l & V-2, Tables V-l to V-4, JA 84-87, 89-103.
These calculations suggest that the danger from lead emissions in the atmosphere is a real one. If substantiated, they warn that air lead concentrations can make a significant contribution to blood lead levels. Two clinical studies, which will be discussed in detail below, have recently provided important empirical support for this theoreti*428cal conclusion. The chamber study, see pages 434 to 435 of 176 U.S.App.D.C., pages 62 to 63 of 541 F.2d infra, analyzed subjects with low dietary lead intake (approximately 100 ug/day, or 10 ug daily absorption) but high respiratory intake (approximately 49-60 ug daily absorption 2). The blood lead level of the experimental subjects went from 20 ug to 37 ug in a four-month period. These results are similar to those predicted by the above calculations at such elevated exposure. Third Health Document at V-8 & Table V-5,.JA 88 & 102. The pilot lead isotope study, see page 434 of 176 U.S.App.D.C., page 62 of 541 F.2d infra, demonstrated that in an atmosphere of 2 ug/m8, 28 percent of a subject’s lead intake was attributable to respiratory lead. This closely corresponds to the calculations made for a “standard” man in such an atmosphere; EPA’s predicted value at 2 ug/m8 was 29 percent. Id. at V-3 & Table V-l, JA 83 & 100. Thus the theoretical calculations, supported by the empirical data, provide substantial support for the Administrator’s conclusion that absorption of airborne lead makes a significant contribution to the body lead burden.
Petitioners’ attack on the Administrator’s analysis is so misguided that it is questionable whether they have even attempted to understand the purpose of these calculations. Instead, they fight again a battle they won over two years ago. Petitioners vigorously attack the Administrator’s calculations as a revitalization of the so-called Goldsmith-Hexter regression equation. Nalco Supp.Br. at 39-40; PPG/duPont Supp.Br. at 32-33. As they rightly observe, EPA abandoned this approach at the time of the Second Health Document. See Second Health Document, Appendix A, JA 214-218. Petitioners do not recognize, however, that the mere fact the Administrator is once again engaged in theoretical calculations does not necessarily mean he has returned to the discredited Goldsmith-Hexter approach.
The Goldsmith-Hexter regression equation was an attempt by EPA to predict a person’s blood lead level as a direct function of the air lead concentration to which he was exposed. This equation would then be. used to find a “safe” air lead concentration to avoid elevated blood lead levels. First Health Document at 3-5 & Table 7, JA 294-296, 305. The approach simply made too many assumptions. In particular, it did not recognize the wide variability of dietary lead intake and thus was of no predictive value. After considering petitioners’ critiques, EPA abandoned its attempt to quantify a safe level of exposure and settled on its current, more qualitative, approach to the evidence. JA 214-218.
The current analysis is much more modest in scope and does not rely at all on the Goldsmith-Hexter data or equation. Well established work has shown that blood lead levels will reach the 40 ug mark if a daily absorption of 60 ug is maintained over time. EPA simply calculates, for various specified situations, the air lead concentrations necessary to produce that daily 60 ug absorption figure. Since petitioners’ objections are not addressed to this use of the Kehoe data, they can hardly be credited.3 On the other hand, use of predictions based on uncertain *429data is plainly within the scope of the precautionary “will endanger” standard. Amoco Oil Co. v. EPA, supra, 163 U.S.App.D.C. at 181, 501 F.2d at 741. And since in at least half the situations considered by EPA an elevated blood lead level would be produced in a “standard” man by airborne lead levels now present in parts of our urban environments, we cannot say the Administrator’s reliance on this data, as bolstered by the clinical studies, was irrational.
2. Epidemiological Studies. The intercity correlations produced by the major epidemiological study, the Seven Cities Study, were negative. The Study was an attempt to correlate atmospheric and body lead levels in seven cities in the United States. No statistically significant correlation was found between the atmospheric lead in a city and the blood lead level of its inhabitants. As the Administrator noted in promulgating the low-lead regulations, residents of New York City had lower average blood lead levels than residents of Philadelphia, despite the fact that somewhat higher airborne lead levels were measured in New York. 38 Fed.Reg. 33735.
Petitioners rely heavily on the negative conclusion of the Seven Cities Study. PPG/ duPont Supp.Br. at 24-25; Nalco Supp.Br. at 34-35; NPRA Supp.Br. at 11-12; Ethyl Supp.Br. at 45-46. The Administrator, however, discounted the significance of the Study’s overall conclusion because of the intercity differences in lead intake from sources other than the air, 38 Fed.Reg. 33735, and instead relied on data in the study relating to individual metropolitan areas. In such areas climate was constant4 and it could more readily be assumed that the lead content of the diet was roughly comparable for all subjects, so that dietary lead’s confounding influence on the data would be minimized. These data showed consistent differences between the blood lead levels of urban and suburban dwellers, with the urban residents, who were exposed to higher air lead concentrations, invariably showing higher blood lead levels. The authors of the Seven Cities Study concluded:
In a given metropolitan area, urban-suburban comparisons tend to minimize the influence of diet and climate. That airborne lead contributes to the relatively higher blood lead concentrations in centercity populations would seem to be the most probable interpretation of this consistent observation.5
JA 892. Upon reanalysis of the underlying data, the Administrator concluded that while air lead was not the most influential factor affecting urban-suburban blood lead levels, it was a significant factor. 38 Fed. Reg. 33735; Third Health Document at V-9, JA 89. Petitioners do not so much attack the Administrator’s reliance on this data as they argue that the intercity comparisons of the Study should be accepted. In this ^context, we cannot say that the Administrator’s statement of reasons for crediting the suburban-urban data while discounting the intercity data was irrational.
For reasons similar to those advanced for discounting the intercity results of the Seven Cities Study, the Administrator also *430gave only slight value to the intercity results of the duPont-sponsored Azar study6 of five groups of 80 subjects — two groups of taxicab drivers and three groups of duPont employees — in four cities around the United States. This study is also strongly relied upon by petitioners to show the lack of a relationship between body and air lead concentrations. Nalco Supp.Br. at 35-36; PPG/duPont Supp.Br. at 26-27; NPRA Supp.Br. at 12. The subjects’ exposure to air lead was monitored continuously for two to four weeks. As in the Seven Cities Study, no significant correlation was found between air lead concentration and body lead burden among all the subjects. Nonetheless, the Administrator noted that by the duPont scientist’s own assessment the “design deficiencies” of the project included
the use of different occupational groups, the use of different cities, data collection during different times of the year, lack of detailed histories and physicals, and perhaps most importantly, the lack of data relative to ingested lead intake.
JA 398-399. See Third Health Document at V-ll, 12, JA 91-92.
While not according significant weight to the study for these reasons, the Administrator did observe a generally positive relationship between average blood lead values and increasing airborne lead exposure. Third Health Document at V-12, JA 92. See JA 396 (Fig. 3). This relationship is especially striking in the data for Los Angeles, the only city in the study in which two groups were evaluated. Taxi drivers exposed to air lead concentrations of 6 ug showed blood lead concentrations some 25 percent higher than office workers exposed to air lead concentrations of 3 ug. JA 392 (Table 1). This finding, again assuming some rough dietary comparability, is consistent with the urban-suburban differences of the Seven Cities Study.7
The inferences the Administrator drew from the Seven Cities Study and the Azar study were bolstered by the preliminary results of a Japanese study8 designed to duplicate the design technique of the Seven Cities Study. Analyzing exposure to air lead concentrations in a country whose concentrations of airborne lead, averaging 0.65 ug/m3, are “far less than those observed in the U.S. cities,” JA 1130 (emphasis added), the Japanese study reached an inconclusive result similar to that of the Seven Cities Study: “With the present level of lead concentration in the air it is difficult to establish any direct relationship between lead levels in the air and such levels in blood and urine.” Id. 1131.
Nonetheless, once again the urban-suburban gradient clearly emerged. Even at that significantly lower air lead concentration, the study was able to find “a significant positive correlation between ambient [lead] level and blood level for workers by area,” id. 1123, and to conclude that the increase of blood levels with urbanization “was statistically significant,” although of small magnitude. Id. 1130. See also id. 1119, 1122-1123. Given the low levels of air lead concentration throughout Japan, the small magnitude is hardly surprising; what is surprising is that the difference is significant at all. It is for this finding that the Administrator cited the study.
Petitioners are quick to point out that the authors observe:
*431[Wjithout monitoring personal exposure, the urban-rural differential in blood lead may not be considered to reflect the differential in air lead concentration.
JA 1124. See Nalco Supp.Br. at 36; Ethyl Supp.Br. at 47. In context, however, the statement does not have the significance petitioners ascribe to it. The authors have just observed that although air lead concentrations are generally lower in rural areas than in the cities, pockets of high and low air lead concentrations may exist in both areas. JA 1123. Thus they simply caution that without firm proof of individual exposure to air lead it is improper to conclude with certainty that the urban-suburban gradient necessarily reflects air lead differentials. JA 1124. As has been observed above, however, inability to prove conclusively cause and effect relationships is an inherent limitation of epidemiological research. The authors’ warning is proper and does not detract at all from a study of this nature. It is significant, however, that immediately after this warning, the authors suggest the proper inference to be drawn from the data, “No doubt an increased respiratory exposure will contribute to an increased intake of lead into the body,” id. 1124, and conclude, “A positive correlation between the lead levels in air and blood was strongly suggested * * *.” Id. 1131. We think petitioners have failed to show that the Administrator’s reliance on this study was improper.
Beyond the Azar and Seven Cities Studies, the only study affirmatively offered by petitioners to disprove the Administrator’s conclusion is the London taxicab study, an epidemiological study of a much smaller group than any of the work discussed above. The study, conducted by Jones et al.,9 compared smoking and nonsmoking day-shift taxi drivers with similar night-shift drivers, and found no significant blood lead differences between them. No attempt was made, however, to measure the exposure of these drivers to airborne lead; the authors assumed the daytime drivers were exposed to more lead since they showed higher carboxyhemoglobin levels, indicating exposure to exhaust fumes. Because of the failure to measure actual air lead levels, as well as to measure smoking intensity.of the smoking drivers, the Administrator found the results of this study unhelpful (although, more than with the other studies cited by petitioners, dietary lead could be assumed to be relatively constant). Third Health Document at V-12, 13, JA 92-93. The validity of the data is particularly questionable in light of its inconsistency with the Azar data, also presented by petitioners, which showed significant differences between blood leads of taxi drivers and office workers in the same city that corresponded to significant differences in measured air lead exposure. We cannot fault the Administrator’s discounting of the London taxicab study in light of the offsetting evidence to which he pointed.
The Seven Cities, Azar, and Japanese studies all were broadly based epidemiological work. Nonetheless, the Administrator found supportive evidence in all of them when he analyzed the data according to regions so that the effects of diet and climate could be, to the extent possible, minimized. Perhaps the most important epidemiological study before him, however, was the Daines study,10 which did not present the problems of the above studies since it was composed of two studies, each of a single neighborhood. In the first neighborhood studied, the researchers measured, over a two-year period, the blood lead levels of black females living 3.7, 38.1, and 121.9 meters from a major highway; air lead levels at these locations were 4.60, 2.41, and 2.24 ug/m3 respectively. The results showed that women living 3.7 meters from the highway, in high air lead concentrations, averaged blood lead levels of 23.1 ug, while those at the greater distances, and lesser air lead concentrations, averaged 17.4 *432and 17.6 ug.11 That is, women living in air lead concentrations of 4.60 ug/m3 had blood lead concentrations 32 percent.higher than comparable women living in air lead concentrations 2.23 to 2.41 ug/m.3 While, as with an epidemological work, there were uncontrolled, and possibly confounding, variables, there were far fewer here than in any of the studies cited above, and the dramatic positive results suggest their effect was minimal.
Petitioners attempt to discredit the Daines study by pointing to the results of the second neighborhood studied. There the researchers studied, for a three-month period, white women living 33.5 and 457 meters from another major highway in air lead concentrations of 1.95 and 1.73 ug/m3; these women showed blood lead levels of 15.7 and 16.1 ug respectively. Petitioners suggest this shows lack of relationship between air lead and blood lead, thus making the Daines results “equivocal,” Nalco Supp.Br. at 36. Because the average blood lead at the greater distance was slightly higher than at the closer distance, petitioners even suggest that the result shows a relationship opposite to that suggested by the Administrator, PPG/duPont Supp.Br. at 31-32. To the contrary, the Administrator found the results of this study completely consistent with the results of the first neighborhood studied, and concluded that the Daines study provided major support for his position.
We cannot fault the Administrator’s analysis. The suggested inverse relationship is nonexistent. The slight difference between 15.7 and 16.1 ug was analyzed by Daines and found to be insignificant. JA 467. We have been shown no reason to believe his analysis was in any way improper. Meanwhile, the data correspond closely to those found in the first neighborhood at distances beyond 100 feet (approximately 30 meters). Since the study found the air lead levels decreased by over 60 percent after the first 100 feet, JA 465, it is not surprising that no significant differences were found in either of the two neighborhoods beyond that distance.12
The Administrator also drew some support from several studies conducted in communities housing lead smelters which, like automobiles, emit lead particulates into the atmosphere. One such study showed positive differences between children living in smelter and nonsmelter communities13 and another quantified decreasing blood lead levels corresponding with increasing distance from the smelter.14 Since neither study measured air lead concentrations, they could do no more than indicate in a qualitative way the relationship between blood lead levels and exposure to lead in air (and in dustfall). Third Health Document at V-13, JA 93. While petitioners argue that findings relating to inhalation of lead from lead smelters cannot be extrapolated to situations involving inhalation of automotive emissions, PPG/duPont Supp.Br. at 33, their arguments lack force given both the limited nature of the Administrator’s use of these studies and various toxicological studies indicating both types of lead to be comparably absorbed in experimental *433animals. 38 Fed.Reg. 33734, 33735; Third Health Document at VI-5, 6, JA 115-116.
These constitute the primary epidemiological studies before the Administrator. None of them attempted to control the important variable of dietary intake; such control would be impossible in a pure epidemiological study since controls destroy the “real life” nature of the experiment. Recognizing that diet is nonetheless a major factor in human lead intake, the Administrator sought data in which dietary lead could be assumed relatively constant among all subjects.15 Thus he relied most heavily on urban-suburban data from the Seven Cities and Japanese studies, on intracity data from the Azar study, and on the Daines and smelter community studies. In this data base, a significant correlation between air lead and blood lead consistently emerged. We cannot find the Administrator’s reliance on these data, and his rejection of the London taxicab study and the intercity results of the Seven Cities and Azar studies, to be arbitrary and capricious.
3. Clinical Studies. In the Albany chamber study16 two groups of volunteer prisoners lived for weeks in a climate-controlled chamber where they were exposed to airborne lead levels of 10.9 ug/m3 and 3.2 ug/m8 for 23 hours a day. While dietary lead was not measured, a control group lived under similar conditions and was fed the same diet. The controls, however, were exposed to air lead concentrations of only 0.2 ug/m3, thus making valid comparisons possible between groups. Blood lead concentrations did not change significantly among the controls during the experiment. JA 593 (Fig. 1). On the other hand, blood lead levels of those exposed to air lead concentrations of 3.2 ug/m3 increased from 19 to 25 over an 11-week period. Even more dramatically, prisoners exposed to daily air lead concentrations of 10.9 ug/m3 registered enormous blood lead increases: over 18 weeks, their average blood lead almost doubled, increasing from about 20 ug to 37 ug. Besides having independent significance, these results correlate closely to the Administrator’s theoretical' predictions of the effect of exposure to such high air lead levels. See pages 427 to 428 of 176 U.S.App.D.C., 55 to 56 of 541 F.2d supra.
Petitioners argue that these results show little since exposure to even this high air lead level (twice that found in even the most polluted urban areas) did not cause the subjects’ blood lead levels to pass the 40 ug warning point. PPG/duPont Supp.Br. at 30-31; Nalco Supp.Br. at 38. This argument misses the point. These subjects were exposed to high air lead concentrations for only 18 weeks; the Administrator, however, is concerned about the effects of exposure over a lifetime. The rapid rise of blood lead among the experimental subjects in this short period shows that air lead is a significant factor in the body lead burden and that the Administrator’s concern may well be warranted. Petitioners’ argument boils down to an objection that the study is flawed because it did not continue until the subjects were poisoned. We can hardly credit such a complaint.
Petitioners also argue that the study is suspect because the lead particulates to which the subjects were exposed were not identical to those emitted from automobiles. PPG/duPont Supp.Br. at 30; Nalco Supp.Br. at 38. It is true that the particles differed somewhat from those found in automobile exhausts. This was an inherent *434limitation in the study since atmospheric lead pollution could not be precisely reproduced in a laboratory, and the authors acknowledged it as such. Doc. 385 at 17. But while the absorption of lead in the experiment “may differ” from that due to exposure to the ambient air, id., the study was, of course, designed to minimize any differences. Thus over 80 percent of the lead particles in the chamber were in the respirable range, JA 584, a figure that compares favorably with accepted data for lead particles in the ambient air, including data produced by petitioners’ own Azar study (79-90 percent). JA 392 (Table 1). See also Third Health Document at II — 8, JA 41. Moreover, the author of the study, when closely questioned about this point by an Ethyl witness after delivering his paper at an international lead conference in Amsterdam in 1972, suggested that the more complex lead particles present in automobile exhausts would nonetheless be absorbed in much the same way as the particles in the experiment. Doc. Ill at 399-401. The Administrator was aware of this limitation on the chamber study, but concluded that the differences in particle size “were sufficiently small so that inferences can be made regarding effects of airborne lead among the general population.” Third Health Document at V-8, JA 88. Thus he considered petitioners’ objection to this study and rejected it; we cannot say that rejection was arbitrary or capricious.
The second clinical study, the pilot lead isotope study,17 was an ambitious attempt to measure precisely a subject’s total intake and output of lead. A subject was placed for 160 days in a controlled metabolic unit where he was extensively monitored. He was fed a low-lead diet and lived in an atmosphere with a lead concentration of 2 ug/m3, a level frequently found in our major cities. His diet included lead isotope tracers so that separation of dietary and respiratory lead was possible. Based on careful measurements of the quantity of isotope-labeled lead in the blood, the amount attributable to respiratory absorption could be derived. The researchers concluded that at 2 ug/m3 air lead concentration the body absorbed approximately 15 ug of lead daily from the atmosphere, so that 28 percent of the subject’s blood lead was a product of airborne lead. JA 705-706. This conclusion was replicated by preliminary results on a second subject. JA 690. Together with the chamber study, this study provides important experimental support for EPA’s theoretical calculations discussed above, see pages 427 to 429 of 176 U.S.App.D.C., pages 55 to 57 of 541 F.2d supra.
Certainly the project proceeds on some rough assumptions and presents' data on only two subjects, as petitioners quickly point out, Nalco Supp.Br. at 38-39; PPG/duPont Supp.Br. at 27-28, and as the researchers freely conceded. JA 683, 684, 686, 688. The limited number of subjects would seem a necessary consequence of the intrusive, as well as confining and time-consuming, nature of the study. And the very magnitude of the air-blood correlation found, and the limited purpose to which the results were put — to support the independent theoretical calculations, Third Health Document at V-3, JA 83 — convince us that the Administrator’s reliance was not improper.
Petitioners also argue that the result is so inconsistent with the other evidence that it must be rejected. PPG/duPont Supp.Br. at 28; NPRA Supp.Br. at 12. To the contrary, the results are remarkably consistent with the evidence the Administrator credited— the Seven Cities and Japanese urban-suburban data, the Azar intracity data, the Daines study, the chamber study. All of these findings indicate a significant contribution of air lead to the body lead burden; a finding of a 28 percent contribution is consistent with that conclusion. The find*435ing is only inconsistent with the data petitioners urged the Administrator to accept but that he, properly we hold, rejected.
The two clinical studies, then, provide important support for the Administrator’s conclusion. They show that exposure to high airborne lead concentrations can lead to rapid increases in blood lead levels, and that respiratory absorption of lead accounts for a significant amount of the body lead burden. These findings are consistent with the theoretical calculations and support the clear inference of the many epidemiological studies.
APPENDIX B
Evidence That Lead Dustfall from Automobiles is Swallowed by Children with Pica.
To support his conclusion that children with pica swallow dust contaminated by lead from automobile emissions, the Administrator relied on studies showing that a persistent fraction of children with lead poisoning had no contact with leaded paint, thereby implicating another source.1 Other work found a small but significant number of children living in good housing with no apparent exposure to lead paint to have, nonetheless, high lead levels, again suggesting another source.2 Meanwhile, there is evidence tying children to leaded dust. A Rochester, New York study found lead-containing dust on the hands of inner city children with elevated blood lead levels and suggested the hand-to-mouth route as a significant source of lead exposure.3 A study in Charleston, South Carolina found that children living in homes near high lead soil concentrations had higher blood levels than children in randomly selected homes.4 Epidemiological work in the smelter community of El Paso, Texas shows that elevated *436blood leads among children decrease with distance from the smelter.5 This gives substantial support to the claim that children can ingest lead-contaminated dust in significant amounts. And a major epidemiological study in Newark6 supports this inference with evidence relating directly to automobile lead. It found children living within 100 feet of heavily traveled roadways had significantly higher blood lead levels than those living beyond that distance. When only children living near highways were considered, the study showed a direct relationship between blood lead level and traffic density; the more heavily traveled the roadway, the higher the blood lead level.7 The plain thrust of all these studies is that the Administrator’s hypothesis is well founded, that children do ingest toxicologically significant amounts of lead dustfall from automobile emissions.
Petitioners attack these studies for various reasons. Primarily, they argue that too many variables were uncontrolled and that the Administrator cannot properly extrapolate data from lead smelter communities to urban areas in general. We have examined these objections and find them to be without merit. The variables that were uncontrolled were uncontrollable and/or were adequately accounted for by the Administrator.8 The use of the lead smelter data, in *437context, was not improper. Animal toxicological studies show that lead-contaminated dust from smelters and from automobiles is comparably absorbed. Third Health Document at VI-5, 6, JA 115-116. Moreover, the studies were used to prove only that lead-contaminated dust is in fact ingested by children; the source of the contamination was irrelevant to this purpose.
We likewise reject the evidence affirmatively presented by petitioners to disprove the dustfall theory. Most of it only proves that leaded paint is the primary cause of lead poisoning in children, something the Administrator had repeatedly acknowledged.9 Petitioners rely primarily on the Ter Haar study,10 produced by an Ethyl scientist. That study was an attempt to measure the actual contributions of lead-based paint and lead dustfall in eight children suspected of having high lead body burdens. The project tested the children’s feces for the presence of a lead isotope, lead-210, more prevalent in fallout dust than in leaded paint. Six of the eight children showed elevated stable lead concentrations in their feces, indicating elevated lead body burdens, but none of the children showed lead-210 concentrations higher than those in a control group of normal children.
Because of this, the authors concluded that the children suspected of elevated lead body burdens did not ingest dust.
The Administrator criticized the study for failing to measure the children’s exposure to lead-210 in their diet or in the air; exposure by either of these routes could have confounded the results. Third Health Document at VI-11,12, JA 121-122. Moreover, while positive results would have shown conclusively that some children ingested lead through dust, these negative results show only that the six children who tested positive had not eaten lead dustfall. Without some reason to believe their exposure to dustfall was typical of all children, extrapolation of the results beyond the sample studied is unwarranted — and is not suggested by the authors.11 We cannot say the Administrator was irrational in refusing to credit the Ter Haar study, particularly in light of the positive, contrary data on which he relied.
. All sections of the Act pertinent to this case were added by the Clean Air Amendments of 1970, Pub.L. 91-604, Dec. 31, 1970, 84 Stat. 1698-1700.
. The new regulations, together with regulations requiring all gasoline refiners to market at least one line of lead-free gasoline, are set out in 40 C.F.R. § 80 (1975). The lead-free regulations serve a different purpose than the low-lead regulations now before us. Leaded gasoline fouls the catalytic converter emission control system developed by the major automobile companies to meet the air pollutant emission standards set by Congress in § 202 of the Clean Air Act. 42 U.S.C. § 1857Í-1. So as to allow implementation of the catalytic converter, the Administrator ordered the marketing of lead-free gasoline pursuant to his authority under § 211(c)(1)(B), the sister section to § 211(c)(1)(A) at issue here. Section 211(c)(1)(B) gives EPA authority to regulate gasoline additives whose emission products “will impair to a significant degree the performance of any emission control device or system which is in general use * * 42 U.S.C. § 1857f-6c(c)(l)(B). The lead-free regulations were approved by this court in Amoco Oil Co. v. EPA, 163 U.S.App.D.C. 162, 501 F.2d 722 (1974).
On March 5, 1975, the Administrator suspended the 1977 statutory standards for automobile emissions of hydrocarbons and carbon monoxide, establishing interim standards for that model year equal to those now in effect. 40 Fed.Reg. 11900. This action was taken to slow the emission reduction schedule and" thereby minimize a potential health hazard posed by the gradual increase in sulfuric acid emissions produced by the catalytic converter. So far as this decision is relevant to this case, it bears noting that the lead-free gasoline regulations will continue in effect and automobiles will continue to be equipped with catalytic converters for the foreseeable future. See note 68 infra.
.The principal studies and symposia over the years on the subject have included Sayers et al, “Experimental Studies on the Effect of Ethyl Gasoline and Its Combustion Products,” Bureau of Mines, 1927; HEW, “Public Health Aspects of Increasing Tetraethyl Lead Content in Motor Fuel,” Public Health Service Pub. No. 712, 1959; Kehoe, “The Metabolism of Lead in Man in Health and Disease,” The Harben Lectures, 1960, JA 500-579; HEW, “Survey of Lead in the Atmosphere of Three Urban Communities,” JA 789-839; HEW, “Symposium on Environmental Lead Contamination,” Public Health Service Pub. No. 1440, 1966, JA 975-*380984; Tepper & Levin, “A Survey of Air and Population Lead Levels in Selected American Communities,” 1972, JA 840-916; National Academy of Sciences, “Airborne Lead in Perspective,” 1972, JA 3C9-362; EPA and Commission of European Communities, Proceedings of International Symposium, “Environmental Health Aspects of Lead,” Luxembourg, 1973, JA 676-677; EPA and National Institute of Environmental Health Services, Conference on “Low-Level Lead Toxicity,” Raleigh, N.C., 1973.
. Some of the data discussed herein speak of micrograms of lead per 100 grams of blood, while other data report micograms per 100 milliliters of blood. Since the density of blood is close to 1.0, these figures are directly comparable. Therefore, lead concentrations, unless otherwise indicated, will hereafter simply be given in micrograms. See National Academy of Sciences Committee on Biologic Effects of Atmospheric Pollutants, Airborne Lead in Perspective 61 n. * (1972) (hereinafter NAS Report).
. Lead in food and water ultimately can be traced to lead in soil and this, of course, is uncontrollable. The NAS Report concluded that “[tjhere is no evidence that the amount of lead in the diets of people has changed substantially since 1940.” NAS Report at 206.
. A list of local laws regarding control and/or removal of lead-based paints can be found in NAS Report at 77.
. This is EPA’s figure, Third Health Document at II-4, JA 37, and Ethyl Corporation (hereinafter Ethyl), alone among the petitioners, contests it. Supplemental brief of petitioner Ethyl Corporation (hereinafter Ethyl Supp. Br.) at 40. We cannot say that EPA’s estimate is unreasonable. It was determined by discounting the NAS Panel’s conclusion that “about 98% of the airborne lead that can be traced to its source comes from combustion of gasoline.” NAS Report at 31. See also id. at 12-13.
. An alternative approach, considered and rejected by EPA, is to trap and remove lead emissions from the exhausts of automobiles using leaded gasoline. See note 66 infra.
. See, e. g„ 116 Cong.Rec. 19207 (1970) (remarks of Rep. Skubitz) (“the Government can require that the oil companies get the lead out”); id. at 19228-19230 (colloquy between Reps. Waggoner & Staggers); id. at 19234 (remarks of Rep. Williams); id. at 19239 (remarks of Rep. Sebelius).
. Regulations promulgated for the latter reason were approved in Amoco Oil Co. v. EPA, supra note 2. See note 2 supra.
. JA 292. This document was subsequently revised by Health Hazards of Lead (Rev. April 11, 1972), JA 254; Atmospheric Lead and Pub*382lic Health (April 11, 1972), JA 276, and Corrections and Additions to Health Hazards of Lead (April 27, 1972), JA 272.
. The reduction would proceed in the following steps:
1.7 g/gal. after Jan. 1, 1975
1.4 g/gal. after Jan. 1, 1976
1.0 g/gal. after Jan. 1, 1977
0.8 g/gal. after Jan. 1, 1978
0.5 g/gal. after Jan. 1, 1979
40 C.F.R. § 80.20 (1975). According to EPA calculations, when the overall lead content of both leaded and unleaded gasoline averages 0.5 grams per gallon, leaded gasoline would contain an average of 1.25 grams of lead per gallon, the same as under the original regulations. See 38 Fed.Reg. at 33739 (1973).
. Commentators have been uniformly critical of the majority opinion. See Gardner, Federal Courts and Agencies: An Audit of the Partnership Books, 75 Colum.L.Rev. 800, 801 & n.77 (1975); Note, Judicial Review of the Facts in Informal Rulemaking: A Proposed Standard, 84 Yale L.J. 1750, 1767-68 & nn.81-82 (1975); Note, Reserve Mining — The Standard of Proof Required to Enjoin an Environmental Hazard to the Public Health, 58 Minn.L.Rev. 893, 918-19 n.116 (1975). See also Reserve Mining Co. v. EPA, 514 F.2d 492, 519-520 (8th Cir. 1975) (en banc).
. When EPA acts under § 211(c)(1)(A) it is essentially telling manufacturers how to make their fuels, a task Congress felt the Agency should enter upon only with trepidation. See, e. g., 116 Cong.Rec. 32920 (1970) (remarks of Sen. Baker); id. at 19229 (remarks of Reps. Rogers & Waggoner). On the other hand, when the Agency acts under § 202, it is only mandating an end product — regulated emissions. The method for achieving the required result is entirely in the hands of the manufacturers.
Nonetheless, deference to regulation under § 202 is not mandatory. The Administrator is only required to “consider” the possibility of regulating under that section instead of under § 211. This language is in sharp contrast to the version of § 211 that was passed by the House. The House version would have allowed regulation under § 211 only after the Administrator made a specific finding “that it is not otherwise technologically or economically feasible to achieve the emission standards established pursuant to section 202 of this Act.” H.R. 17255, 91st Cong., 2d Sess. § 210(g)(1) (1970). This mandatory deference to § 202 regulation was removed in conference and the present more flexible language was substituted. See pages 394-395 of' 176 U.S.App.D.C., pages 22-23 of 541 F.2d infra.
. At oral argument, petitioners claimed the regulations were void because the Administrator had failed to couch his ultimate finding in the language of the statute itself. See also supplemental brief of petitioner National Petroleum Refiners Association (hereinafter NPRA Supp.Br.) at 5. The short answer to the argument is that petitioners’ view of the facts is erroneous. While interpreting the “will endanger” standard to mean “presents a significant risk of harm,” 33 Fed.Reg. 33734, the Administrator ultimately did make his finding in the language of the statute:
These regulations are based upon a determination by the Administrator that the emission product of a fuel or additive will endanger the public health * * *.
Id. at 33741. See 40 C.F.R. § 80.1 (1974).
In any case, however, the issue is spurious. It is well established that ultimate findings do not have to be expressed at all, let alone be expressed in the language of the statute. Rather, absent compelling countervailing considerations, an ultimate finding will be implied from the action taken. Pacific States Box & Basket Co. v. White, 296 U.S. 176, 186, 56 S.Ct. 159, 163, 80 L.Ed. 138, 146 (1935); Martin v. Mott, 25 U.S. (12 Wheat.) 19, 32-33, 6 L.Ed. 537, 30-32 (1827) (Story, J.). Cf. Joseph v. FCC, 131 U.S.App.D.C. 207, 211-212, 404 F.2d 207, 211-212 (1968). The law is fully developed in 2 K. Davis, Administrative Law Treatise § 16.07 at 455-59 (1958).
. We note that even if we did not agree fully with the Administrator’s interpretation of the Act, we would be obliged to accord it considerable deference. As the Supreme Court recently held in approving, despite the contrary views of several circuits, the Administrator’s reading of another section of the Clean Air Act:
Without going so far as to hold that the Agency’s construction of the Act was the only one it permissibly could have adopted, we conclude that it was at the very least sufficiently reasonable it should have been accepted by the reviewing courts.
Train v. Natural Resources Defense Council, Inc., 421 U.S. 60, 75, 95 S.Ct. 1470, 1479, 43 L.Ed.2d 731, 743 (1975). See also id. at 87, 95 S.Ct. at 1485, 43 L.Ed.2d at 750.
. It is linguistically clear, of course, that one can be “endangered” without actually being harmed. Nonetheless, some risk of harm is necessary. State v. Fine, 324 Mo. 194, 23 S.W.2d 7, 9 (1929). Webster defines “endanger” as “to bring into danger or peril of probable harm or loss.” Webster’s Third New International Dictionary 748 (1961) (emphasis added).
Not all courts have agreed that “probability” of harm is the proper determinant of danger. Where grounds for divorce are few, for instance, courts have interpreted laws allowing divorce because of inhuman treatment that “endangers the life” to require only the slightest possibility of actual loss of life. Thus action that endangers health has been held to endanger life on the theory that the former cannot be endangered without threatening the latter. See Cole v. Cole, 23 Iowa 433, 438 (1867); Beebe v. Beebe, 10 Iowa 133, 139 (1859). In one of the latest of these cases, mistreatment that deprived a spouse of needed rest and peace of mind was held to “endanger” life. Smith v. Smith, 258 Iowa 557, 561, 138 N.W.2d 453, 456 (1966). We need not decide here how remote the possibility of actual harm could be under the “will endanger” standard since we accept the Administrator’s determination in this case that a “significant” risk to health falls within the statutory language. See pages 389-392 of 176 U.S.App.D.C., pages 17-20 of 541 F.2d infra.
. Petitioner Ethyl suggests that while these may indeed be examples of endangerment they differ from the threat from automotive lead emissions in that
plagues do cause death and illness; violent storms do cause damage — known facts that may be experienced by the threatened community.
Ethyl Supp.Br. at 11 (footnote omitted). We may preliminarily observe that the absorption of lead does cause lead poisoning, a known fact that may be evaluated by the public and the EPA. However, in so far-as Ethyl is complaining that the mechanism by which plagues and storms cause damage is well known while the question of the relation between lead automobile emissions and the absorption of lead is less certain, Ethyl’s observation only supports the reading of § 211(c)(1)(A) as a precautionary statute. The massive diffusion of airborne lead is a gross environmental modification never before experienced. Of course, there are no past disasters of the kind anticipated by the Administrator on which the community’s experience may be based. This, however, is inherent in such a threat and does not imply that no danger is posed by it. We believe the precautionary language of the Act indicates quite plainly Congress’ intent that regulation should precede any threatened, albeit unprecedented, disaster. Ethyl is correct that we have not had the opportunity to learn from the consequences of an environmental overdose of lead emissions; Congress, however, sought to spare us that communal experience by enacting § 211(c)(1)(A).
.Since Congress seemed to assume that the meaning of the threshold determination required by § 211(c)(1)(A) would be self-evident, there is a complete absence of helpful legislative history. What little there is of relevance is discussed at pages 393-395 of 176 U.S.App.D.C., at pages 21-23 of 541 F.2d infra. See also note 89 infra.
. Should there be any doubt about this conclusion, it is quickly resolved by reference to § 109. There Congress made it quite clear how it would refer to anything other than actual harm. Section 109(b)(2) provides that national secondary ambient air quality standards for pollutants listed under § 108 be prescribed with a margin of safety; they should be sufficient to protect against “any known or anticipated adverse effects associated with the presence of such air pollutant in the ambient air.” 42 U.S.C. § 1857c-4(b)(2) (emphasis added). When Congress uses the phrase in § 108 without the modifier “known or anticipated” it plainly embraces only the usual meaning of adverse effects, e., known adverse effects or actual harm.
. The Administrator appears to have a measure of discretion in determining whether to list a pollutant under § 108, which, by its terms, speaks of the exercise of his “judgment.” See 38 Fed.Reg. 33740 (1973). Amicus Natural Resources Defense Council (NRDC) has argued that listing of air pollutants under § 108 is mandatory. Its petition for review of these regulations raising that issue was dismissed by the division for want of jurisdiction, Natural Resources Defense Council, Inc. v. EPA, No. 74-1023 (D.C.Cir. September 11, 1974), and NRDC has since brought a successful citizen’s suit under § 304 of the Act, 42 U.S.C. § 1857h-2, to compel issuance of national ambient air standards for lead. Natural Resources Defense Council, Inc. v. Train, 411 F.Supp. 864 (S.D.N.Y. 1976).
. Thus we must reject amicus NRDC’s suggestion that the standards for action under §§ 108 and 211 are the same.
. Petitioners’ arguments that the threshold determination under the “adverse effect” standard is less rigorous than under the “will endanger” standard are either spurious or misdirected. Nalco Chemical Company (Nalco) and Ethyl argue that the presence of the phrase “in his judgment” in § 108, and its absence in § 211, means the Administrator has greater discretionary power under the former section. Ethyl Supp.Br. at 21; Nalco Supp.Br. at 22. As we shall explain below, see note 37 infra, the Administrator retains the same (if not greater, see note 21 supra) discretionary power under § 211.
Nalco points to the “margin of safety” language of § 109 as proof of a “lower standard” under § 108, Nalco Supp.Br. at 22, without recognizing that the margin of safety refers only to the implementing requirement of formulating standards and not to the threshold decision to regulate. NPRA likewise confuses this point by suggesting that the listing requirement of § 108 is like the reporting requirements of § 211(a), so that the “margin of safety” language is comparable to § 211(c)(1)(A), only more generous. NPRA Supp.Br. at 42-43. This ignores the fact that once a pollutant is listed under § 108 the decision to regulate is made; standards under § 109 must follow. On the other hand, additives reported under § 211(a) are not necessarily regulated; regulations are premised only on a § 211(c)(1)(A) finding of endangerment. Thus, as suggested in text, the threshold determination under § 108 is properly compared to the threshold determination under § 211(c)(1)(A).
Ethyl argues that an “adverse effect” does not have as severe a connotation as “endangerment.” Ethyl Supp.Br. at 19 n.27. Even if true, however, this argument has nothing to do with whether the threshold determination to regulate is, or is not, precautionary. Even if § 108 allowed regulation of less severe effects than does § 211, regulations could still be premised only on a finding of an actual effect, while § 211 regulations could still be premised on a precautionary finding of threatened, albeit more severe, harm.
.In essence, petitioners are suggesting § 202 is more properly read to provide for regulation of any automotive emission
which in his judgment causes or contributes to, or is likely to cause or contribute to, air pollution which causes or contributes to, or is likely to cause or contribute to, the endangerment of the public health or welfare.
Quite obviously, this is not what Congress said.
. Section 211 simply skips this chain of causation and requires instead that the emission products of the fuel additive to be regulated endanger the public health. This omission of a step does not support petitioners’ “in and of itself” theory, see pages 401-403 of 176 U.S.App.D.C., pages 29-31 of 541 F.2d infra, but rather is responsive to the different intendments of §§ 101 and 211. See note 27 infra. In any case, it is plain that for regulation under § 211 the emission products must directly (although not necessarily by themselves) endanger the public health; whether they contribute to air pollution that in turn endangers the public health, as required by § 202, is irrelevant. Since the Administrator found that lead emissions directly endanger the public health, whatever leniency § 202 provides is irrelevant to this case.
Nalco takes issue with the dismissal of this question and asks sarcastically, “If causation is irrelevant to the regulation of fuel additives, why is EPA concerned about meeting any standard at all in these regulations.” Nalco Supp.Br. at 20-21. Nalco is confusing two separate causation issues. There is undoubtedly a causation issue about whether lead emissions cause a danger to public health. On the other hand, however, there is no causation issue about whether lead emissions contribute to lead air pollution. There is no such issue, first, because § 211, unlike § 202, does not demand such a finding and, second, because in any case petitioners could not, and do not, contest the validity of that assertion. It is only this issue of causation, to which the “likely” language of § 202 relates, that EPA argues, and we agree, is irrelevant to this case. See EPA Supp.Br. at 17-18.
. Thus while Congress preferred emission regulation under § 202 to fuel content regulation under § 211, see note 14 supra, there is no reason to assume, as NPRA argues, NPRA Supp.Br. at 31-35, that in situations where § 211 regulation is proper, t.he Administrator must find greater potential harm before acting. Cf. note 23 supra.
. NPRA recognizes that a literal reading of § 202 produces the result suggested above. Thus it argues that this result is “meaningless, or at best tautological,” NPRA Supp.Br. at 32, so that its “likely” danger theory emerges as an acceptable, although linguistically incorrect, alternative. Id. at 33. NPRA argues that the above reading would require
the Administrator to determine whether “the emission of any air pollutant ... is likely to cause or to contribute to, air pollution . . ..”
Id. at 32 (emphasis in original). Since under this reading, NPRA argues, the Administrator would always reach a positive conclusion, this reading must be incorrect. We agree that this reading is of little value, but we do not think it is the proper result of our analysis above. Rather, we think that to regulate under § 202 the Administrator must find that emission of the air pollutant is likely to cause or contribute to dangerous air pollution. This addition is important, for not all air pollutants contribute to dangerous air pollution and, more importantly, not all dangerous air pollution is caused by air pollutants that are, themselves, dangerous. Thus hydrocarbons, whose emission is regulated by § 202, are not themselves always dangerous, but are properly regulated because they react in sunlight to form smog, which is dangerous. See S.Rep. 89-192, 89th Cong., 1st Sess. 5-6 (1965); EPA Supp.Br. at 18 n.15. Thus, far from stating a tautology, § 202 allows for the regulation of such apparently innocent pollutants, which indirectly cause dangerous pollution.
. See note 17 supra. See also Environmental Defense Fund, Inc. v. EPA, 150 U.S.App.D.C. 348, 465 F.2d 528 (1972), where in interpreting *389the more rigorous statutory language “imminent hazard” which must be found before the registration for a pesticide may be suspended pending the conclusion of cancellation proceedings, 7 U.S.C. § 136d(c) (Supp. II 1972), we concluded, per Judge Leventhal:
It is enough if there is substantial likelihood that serious harm will be experienced during the year or two required in any realistic projection of the administrative process.
Id. at 360, 465 F.2d at 540 (emphasis added).
In another case interpreting the standards for cancellation of a pesticide under the same statute, we held, per Judge Wilkey, that a showing of “potentially great dangers from DDT” sufficed as a basis for cancellation. Environmental Defense Fund, Inc. v. EPA (Coahoma), 160 U.S.App.D.C. 123, 128, 489 F.2d 1247, 1252 (1973).
. Cf. Environmental Education Act, 20 U.S.C. § 1531 et seq.; Environmental Quality Improvement Act of 1970, 42 U.S.C. § 4371 et seq.; National Environmental Policy Act of 1969 (NEPA), 42 U.S.C. § 4321 et seq. While initially the procedural requirements of NEPA consumed judicial energies most conspicuously, see e. g., Natural Resources Defense Council v. Morton, 148 U.S.App.D.C. 5, 458 F.2d 827 (1972); Calvert Cliffs’ Coordinating Committee, Inc. v. AEC, 146 U.S.App.D.C. 33, 449 F.2d 1109 (1971), courts are increasingly according substantive weight to the NEPA language. See Sierra Club v. Morton, 169 U.S.App.D.C. 20, 37-39 & n.25, 514 F.2d 856, 873-875 & n.25 (1975), and cases cited therein, cert. granted, sub nom. Kleppe v. Sierra Club, 423 U.S. 1047, 96 S.Ct. 772, 46 L.Ed.2d 635 (1976).
. See Green, The Risk-Benefit Calculus in Safety Determinations, 43 Geo.Wash.L.Rev. 791 (1975); Handler, A Rebuttal: The Need for a Sufficient Scientific Base for Government Regulation, id. at 808. Both authors agree that government safety determinations should be preventive and based on assessment of risks. Dr. Handler differs from Professor Green in arguing that risks should be quantified before regulatory decisions are made. Professor Green believes that quantification is not always necessary or possible, and that the public health is better served by the making of value judgments, however inexact.
. See note 17 supra.
. This proposition must be confined to reasonable limits, however. In Carolina Environmental Study Group v. United States, 166 U.S.App.D.C. 416, 510 F.2d 796 (1975), a division of this court found the possibility of a Class 9 nuclear reactor disaster, a disaster of ultimate severity and horrible consequences, to be so low that the Atomic Energy Commission’s minimal consideration of the effects of such a disaster in an environmental impact statement prepared for a new reactor was sufficient. Likewise, even the absolute certainty of de minimis harm might not justify government action. Under § 211 the threatened harm must be sufficiently significant to justify health-based regulation of national impact. Ultimately, of course, whether a particular combination of slight risk and great harm, or great risk and slight harm, constitutes a danger must depend on the facts of each case.
. Nalco devotes several pages of its brief to arguing that there is a distinction between “risk” and “danger” that EPA fails to recognize. Nalco Supp.Br. at 9-11, citing Reserve Mining Co. v. EPA, supra note 13, and Power Reactor Development Co. v. International Union of Electricians, 367 U.S. 396, 81 S.Ct. 1529, 6 L.Ed.2d 924 (1961). To the extent Nalco argues that risk and danger are not synonymous, Nalco battles with a straw man. EPA recognizes that a risk to public health is not necessarily a danger to public health; it only argues that a significant risk of widespread lead poisoning constitutes such a danger. To the extent Nalco argues there is no element of risk in danger, it is plainly wrong, as the cases it cites demonstrate. In support of its position, Nalco quotes the following language from Reserve Mining, in which the en banc court compares its present opinion with the “stay” opinion of a division of the court, Reserve Mining Co. v. United States, 498 F.2d 1073 (8th Cir. 1974), in which the division stayed the District Court’s injunction ordering the immediate closing of the Reserve Mining plant:
As will be evident from the discussion that follows, we adhere to our preliminary assessment that the evidence is insufficient to support the kind of demonstrable danger to the public health that would justify the immediate closing of Reserve’s operations. We now address the basic question of whether the discharges pose any risk to public health and, if so, whether the risk is one which is legally cognizable.
514 F.2d at 507. The Eighth Circuit is not drawing a sharp line between “danger” and “risk,” but is only contrasting degrees of danger. Nalco fails to note that the “legally cognizable” risk ultimately found was that Reserve’s discharges were “endangering” the public health within the meaning of the Federal Water Pollution Control Act. See page 391 of 176 U.S.App.D.C., page 19 of 541 F.2d infra. Necessarily, “risk” is an element of “danger,” and the Reserve Mining court explicitly recognized that fact. 514 F.2d at 520.
The Power Reactor case is likewise of no support to Nalco. There the Supreme Court simply recognized that when regulations juxtaposed, and ordered, findings of “undue risk” and “endanger,” an “undue risk” was not intended to mean “endanger.” The Court did not say that risk was not an element of danger, only that in the regulations before it an “undue” risk was not necessarily a sufficient risk to constitute a “danger.”
. Reserve Mining involved issues not only under FWPCA, but under § 407 of the Rivers and Harbors Act of 1899, 33 U.S.C. § 401 et seq., the federal common law of public nuisance, and various Minnesota air and water pollution laws, Minn.Stat.Ann. § 116.081(1); id. § 115.-07(1); id. § 105.41. Finding the evidence of *391danger posed by Reserve Mining’s waste emissions into the air more compelling than that of the danger posed by the water emissions (since a correlation between inhalation, but not ingestion, of asbestos fibers and cancer could be found), the Eighth Circuit found the continued air emissions to be in violation of various Minnesota regulations and ordered Reserve to “promptly take all steps necessary to comply” with the law. Reserve Mining Co. v. EPA, supra note 13, 514 F.2d at 538. With respect to the water pollution, as described in text, the court found the probabilities of danger to be
low for they do not rest on a history of past health harm attributable to ingestion but on a medical theory implicating the ingestion of asbestos fibers as a causative factor in increasing the rates of gastrointestinal cancer among asbestos workers.
Id. at 536 (emphasis added). Thus the court only ordered cessation of dumping within a “reasonable time.” Id. at 538.
. Petitioners properly point out that, unlike lead, there is no known safe human exposure level for carcinogens. Nalco Supp.Br. at 9; Ethyl Supp.Br. at 29. See Society of the Plastics Industry, Inc. v. OSHA, 509 F.2d 1301, 1307 (2d Cir.), cert. denied, 421 U.S. 992, 95 S.Ct. 1998, 44 L.Ed.2d 482 (1975); Industrial Union Department, AFL-CIO v. Hodgson, 162 U.S.App.D.C. 331, 499 F.2d 467 (1974). This does not, however, imply that the harm caused by lead poisoning is less significant than that caused by cancer, only that safe human exposure levels to the causes of the two diseases may differ. This the Administrator recognized. Unlike the actions of the Secretary of Labor in the above cited cases, he did not order the lead content of gasoline reduced to the lowest detectable levels, but rather directed a phased cutback to what he deemed a safe level.
In any case, however, even if cancer is considered more serious than lead poisoning, the Administrator acted based on an assessment that the risk of lead poisoning from automobile emissions was considerably greater than the cancer risk that motivated the Reserve Mining court. Thus this greater risk of an arguably lesser harm still constitutes endangerment.
. This conclusion that a “significant risk of harm” states a sufficient probability of occurrence to fall within the “will endanger” standard is bolstered by the fact that Congress did not employ various modifiers frequently used (in the Clean Air Act and elsewhere) to mandate more certain endangerment. Thus Congress did not require that the lead emissions “clearly endanger” the public health, cf. 21 U.S.C. § 454(c); id § 661(c); 50 U.S.C. § 1517, or that the emissions pose an “imminent and substantial endangerment” to the public health, cf. 33 U.S.C. § 1364 (Supp. IV 1974); 42 U.S.C. § 300i(a); id. § 1857c-10(b); id. § 1857c-7(c)(1). See Reserve Mining Co. v. EPA, supra note 13, 514 F.2d at 528. We find these omissions more significant than the use of the modifier “will,” which Ethyl argues adds “a particular degree of certainty to the probability that must be shown.” Ethyl Supp.Br. at 12. To the contrary, “will” only makes it clear that the standard is one of danger and nothing less, a conclusion with which we agree.
Whether the evidence relied upon by the Administrator is sufficient to support his finding of a “significant risk of harm” is discussed below at pp. 405-420 of 176 U.S.App.D.C., at pp. 33-48 of 541 F.2d infra.
. Besides the arguments discussed below, petitioners support their claim that a “factual” not judgmental decision is required by pointing to the absence in § 211 of a phrase expressly allowing the Administrator to use his “judgment,” in contrast to the presence of such a phrase in §§ 108 and 202. Nalco Supp.Br. at 21-22; Ethyl Supp.Br. at 21-22. The argument ignores an important difference between the cited sections and § 211. Sections 108 and 202 are mandatory in their terms; under both sections the Administrator “shall” regulate if “in his judgment” the pollutants warrant regulation. Because of the mandatory nature of the provisions, express provision for administrative discretion via the “judgment” phrase is necessary. By contrast, § 211 is permissive; the Administrator “may” regulate if emissions “will endanger” the public health. Since discretion is provided in the directive to the Administrator, the safety valve of a “judgment” phrase is unnecessary. On the other hand, since we find the exercise of judgment to be implicit in a determination of “danger,” there is no need for an express statement of that power and no reason to ascribe significance to the omission of the phrase.
The irrelevancy of the omission is shown by the Amoco decision, where the court construed § 211(c)(1)(B), which, like its sister section at issue here, does not contain the phrase “in his judgment.” Nonetheless, the court recognized *393the Administrator must, necessarily, have the power to assess risks and make policy decisions under that section whenever the determinations called for are judgmental. Amoco Oil Co. v. EPA, supra note 2, 163 U.S.App.D.C. at 180-181, 501 F.2d at 740-741, quoted in text at p. 395 of 176 U.S.App.D.C., at p. 23 of 541 F.2d infra.
. Petitioner Nalco argues that if the specific requirement, and its subsequent deletion, have substantive effect at all, that effect relates not to the threshold decision to regulate, but to the subsequent implementing decisions about how to regulate. Nalco Supp.Br. at 17-20. It is true that a literal reading of the House version, proposed § 210(g)(1), suggests that specific findings are required in setting “standards” for fuel additives rather than for the threshold decision to regulate. Nonetheless, it is doubtful that the language was intended to exempt the threshold decision to regulate from the specific findings requirement, and equally doubtful that a court would ever have so interpreted it. First, as a practical matter it is questionable whether specific findings on “standard” setting could be made without a specific finding as to danger. The threshold finding would seem to be a prerequisite to any subsequent specific findings. Second, the legislative history makes it clear that the House intended the specific findings requirement to apply to the threshold regulatory decision. In presenting the proposed bill to that chamber Rep. Staggers, chairman of the committee that drafted the bill, was confronted with questions about the basis on which the Administrator could regulate fuel additives. Rep. Staggers responded that the Administrator could regulate only “[i]f he has the facts, and he has proven this by facts, that they are a danger and poisonous * * 116 Cong.Rec. 19229 (1970). To confirm that regulation could proceed only upon a threshold factual determination of danger, Rep. Staggers then quoted the specific findings provision that Nalco now argues is inapplicable to this question. Id. at 19230. Likewise, Rep. Rogers, a member of Rep. Staggers’ committee, made clear that the threshold determination — that the gasoline additive endangers health — must be made on the basis of facts and findings. Id. at 19231. These are the only specific comments on proposed § 210(g)(1), and as they are made by knowledgeable committee members the legislative history is clear that specific findings were intended for the threshold decision to regulate.
This conclusion is bolstered by § 211(c)(1)(B) and its interpretation by the Amoco court. Section 211(c)(1)(B), unlike § 211(c)(1)(A), retains the finding requirement and the requirement is phrased as ambiguously as the specific findings requirement in the House version of § 211(c)(1)(A). The Amoco court recognized that a literal reading of this findings requirement (which it termed “awkwardly drafted,” Amoco Oil Co. v. EPA, supra note 2, 163 U.S.App.D.C. at 179, 501 F.2d at 739), would produce an anomalous result, just as the literal reading now urged by Nalco of the deleted findings requirement would produce an anomalous result. “Thus in a literal sense the provision requires ‘findings with respect to’ the actual items of data which the Administrator must ‘consider.’ ” Id., 163 U.S.App.D.C. at 176, 501 F.2d at 736 (emphasis in original). The court rejected this literal, nonsensical, reading and instead read the statute to require a threshold factual finding that regulation was necessary under § 211(c)(1)(B). Id., 163 U.S.App.D.C. at 176-178, 501 F.2d at 736-738.
Based on the legislative history of the House proposed § 210(g) and the Amoco precedent relating to the parallel section, it is clear that had the proposed specific findings requirement been enacted into law, it would have been interpreted, as suggested in text, as requiring a specific factual finding for the threshold decision to regulate, and not necessarily specific findings for all the subsequent implementing decisions.
.The original Senate version of the bill provided:
The Secretary may from time to time on the basis of information obtained under subsection (b) of this section [which required fuel manufacturers to furnish various information to the Secretary] or other information available to him, by regulation control or prohibit the introduction into commerce of any fuel or fuels for use in vehicle engines if the combustion or evaporation of such fuel produces emissions which endanger the public health or welfare * * *.
S. 4358, § 212(c)(1), 91st Cong., 2d Sess. (1970). The Senate bill drew a distinction between regulation on public health and public welfare grounds and required that regulation for the latter reason be preceded by public hearings. Id. § 212(c)(2). With minor verbal changes and deletion of the health-welfare dichotomy the Senate version became § 211(c)(1).
Petitioners argue that the Senate as well as the House contemplated a factual threshold determination. Nalco Supp.Br. at 15-16; Ethyl Supp.Br. at 24-26; NPRA Supp.Br. at 30. Their arguments are without merit. While the House debate contains several references to the need for such a determination, always coupled with discussion of the “specific findings” requirement, see notes 38 supra and 40 infra, petitioners can point to no express reference in the Senate proceedings to the need for a factual threshold finding. Instead they identify discussions of danger and suggest that these somehow imply that the assessment of danger be factually based. The implication is nonsensical; the cited discussions do no more than reiterate the language of the statute, which does not in terms require a factual finding. See S.Rep. No. 91-1196, 91st Cong., 2d Sess. 117 (1970) (“if the combustion or evaporation of such fuel produces emissions which endanger the public health or welfare”); id. at 33-34 (“emission that is a direct endangerment to the public health”) (see notes 25 & 36 supra); 116 Cong.Rec. 32921 (1970) (statement of Sen. Baker) (“emissions that, in and of themselves, endanger the public health or welfare”) (see page — of 176 U.S.App.D.C., pages 29-30 of 541 F. 2d infra). When compared with the explicit House discussion of the need for a factual finding, these statements only show that the Senate intended no such thing. Ethyl also cites this excerpt from the Senate Report:
“The [Administrator] may prohibit the use of any fuel in commerce which may provide emissions that, he finds, would endanger the public health.” Sen.Rep. No. 91-1196, 91st Cong., 2d Sess. 64 (1970).
Ethyl Supp.Br. at 25 (emphasis in original). Since the Senate bill explicitly deleted the finding requirement for action under § 211(c)(1)(A), this language can hardly be taken as supporting its sub silentio inclusion. If anything, the language, which uses “finds” in its colloquial, not legal, sense, only demonstrates the judgmental nature of the Administrator’s decision. See note 37 supra.
. For this reason, statements made in the House about the bill should be used only with great care in any attempt to assess legislative intent. House discussion centered on a significantly more rigorous bill, and statements such as those offered by petitioners from the House debate prove only the effect of the ultimate deletion of the “specific findings” requirement. See Ethyl Supp.Br. at 24; Nalco Supp.Br. at 15; NPRA Supp.Br. at 30. Rep. Staggers’ statement, for instance; directly relates the need for a factual threshold determination to the “specific findings” requirement. As such, it is strong support for the conclusion that deletion of the requirement was intended to do away with such a rigorous threshold requirement. See note 38 supra.
. Neither the House managers’ conference report nor the summary of the conference agreement prepared for the Senate address directly the effect of the changes here at issue. See Conference Agreement on the Clean Air Amendments of 1970, H.R.Rep. No. 91-1783, 91st Cong., 2d Sess. 52-53 (1970) U.S.Code Cong. & Admin.News 1970, p. 5374; Summary of the Provisions of Conference Agreement on the Clean Air Amendments of 1970, 116 Cong. Rec. 42384, 42385-42386 (1970). The most relevant discussion, which is only oblique, is in the Senate conference report, where the need for flexibility in applying § 211 is made clear:
[T]he conference committee wishes to call the attention of the Administrator to the broad environmental, esthetic and health considerations underlying the enactment of this legislation which should be kept in mind in making these determinations [to control or prohibit a fuel or fuel additive].
116 Cong.Rec. 42386 (1970). See Amoco Oil Co. v. EPA, supra note 2, 163 U.S.App.D.C. at 173-174, 501 F.2d at 733-734. In the absence of legislative history discussing the change, we must rely on the language of the present law and its contrast to the rejected House version.
. Section 211(c)(2)(C), 42 U.S.C. § 1857f-6c(c)(2)(C), quoted in text at p. 404 of 176 U.S.App.D.C., at p. 32 of 541 F.2d infra.
. See note 14 supra.
. The House bill required the Administrator to act only on the basis of the required specific findings. The Senate and final bill allows him to act on the basis of information obtained under § 211(b), which requires fuel and fuel additive manufacturers to furnish the Administrator with information about their products’ content and, at his request, to conduct various tests on the compounds and furnish him with descriptions of testing techniques. The Administrator is also free to act on the basis of any “other information available to him,” § 211(c)(1) (emphasis added), thus making the basis for action under § 211(c)(1)(A) unlimited.
.The opinion refers to § 4(c), but § 4(b), which requires the agency to “incorporate in any rules adopted a concise general statement of their basis and purpose,” was obviously intended. See § 4 of the APA, 60 Stat. 239 (1946), 5 U.S.C. § 553.
. Clean Air Act, 42 U.S.C. § 1857 et seq.; Federal Water Pollution Control Act, 33 U.S.C. § 1151 et seq.; Federal Insecticide, Fungicide, and Rodenticide Act, 7 U.S.C. § 135 et seq.
. Reserve Mining Co. v. EPA, supra note 13.
. Society of the Plastics Industry, Inc. v. OSHA, supra note 35.
. Industrial Union Department, AFL-CIO v. Hodgson, supra note 35.
. Environmental Defense Fund, Inc. v. EPA (Coahoma), supra note 28.
. Environmental Defense Fund, Inc. v. EPA (Shell), 167 U.S.App.D.C. 71, 510 F.2d 1292 (1975).
. Even scientific “facts” are not certain, but only theories with high probabilities of validity. Scientists typically speak not of certainty, but of probability; they are trained to act on probabilities that statistically constitute “certainties.” See generally T. Kuhn, The Structure of Scientific Revolutions. While awaiting such statistical certainty may constitute the typical mode of scientific behavior, its appropriateness is questionable in environmental medicine, where regulators seek to prevent harm that often cannot be labeled “certain” until after it occurs. See note 58 infra.
The uncertainty of scientific fact parallels the uncertainty of all fact. In a metaphysical sense, at least, facts are themselves nothing more than risks, or statistical probabilities. See D. Hume, A Treatise of Human Nature, bk. I, pt. Ill, § 6, at 87 (L.A. Selby-Bigge ed. 1958).
. Thus this court, per Judge Wilkey, affirmed EPA’s general ban on the use of DDT even though the evidence was considerably less than certain;
[T]here is a great mass of often inconsistent evidence which was developed at the hearing; this evidence is substantial enough to support the conclusions of the Administrator, although it possibly might support contrary conclusions as well. Considering the evidence in the record as a whole, we cannot say that the Administrator’s decision was not based on substantial evidence, even if the hazardous nature of DDT has not been proved beyond a reasonable doubt. Sufficient evidence has been adduced to show potentially great dangers from DDT, and the Administrator’s decision to cancel the DDT registrations is well within his statutory authority.
Environmental Defense Fund, Inc. v. EPA (Coahoma), supra note 28, 160 U.S.App.D.C. at 128, 489 F.2d at 1252 (emphasis added).
. According to the NAS Panel, present air lead concentrations, which over the largest American cities are 2,000 times greater than air lead concentrations over the mid-Pacific Ocean, have existed for 15 years. NAS Report at 205, JA 356. The Panel attributed these high concentrations primarily to automotive emissions. Id.
. It is for these reasons that, although lead additives have been used for over 50 years, the danger posed by lead emissions is still a question “on the frontiers of scientific knowledge.” Industrial Union Department, AFL-CIO v. Hodgson, supra note 49, 162 U.S.App.D.C. at 338, 499 F.2d at 474. See note 97 infra.
Where existing methodology or research in a new area of regulation is deficient, the agency necessarily enjoys broad discretion to attempt to formulate a solution to the best of its ability on the basis of available information, Permian Basin Area Rate Cases, 390 U.S. 747, 811, 88 S.Ct. 1344, 20 L.Ed.2d 312 (1968).
. Ethyl argues that Industrial Union and Society of the Plastics Industry are distinguishable from this case in that under OSHA the Secretary is directed to set standards for industrial exposure to toxic substances, 29 U.S.C. § 655(b)(5), while action under § 211 is discretionary with the Administrator. Ethyl Supp.Br. at 28-30. Ethyl has merely identified the source of the flexibility granted the Secretary, while failing to set the cited cases apart from this one. Under the Clean Air Act the Administrator’s flexibility is derived not from a command to act, but from a precautionary statute that necessarily includes risk assessment if its preventive purpose is to be achieved. Since there is reason to accord flexibility to the regulator under both acts, the cited cases are good support for the way in which that flexibility is to be exercised.
Nalco asserts that the cases are distinguishable in that Industrial Union and Society of the Plastics Industry involve the manner in which standards are set under OSHA, rather than the Secretary’s power to set standards at all. Nalco Supp.Br. at 12. This purported distinction is no more than a factual difference between the cases. Flexibility is necessary under OSHA only in the manner of setting standards, since the command to set standards is clear. Under the Clean Air Act, on the other hand, flexibility is necessary both in standard-setting and in deciding whether to regulate. The teaching of the cases is nonetheless applicable.
Reserve Mining shows the spurious nature of both asserted distinctions. In that case, as here, there was no command to regulate, only a statutory term of a precautionary nature — “endangering” — to justify flexible decision-making (in Reserve Mining by a court rather than by an *400agency). Likewise, that case involved not only the manner of standard-setting, but the threshold decision to regulate as well. Nonetheless, Reserve Mining is fully consistent with Industrial Union and Society of the Plastics Industry, and all three cases support our conclusion here.
. Or, as with OSHA, mandatory in its command to act. See note 56 supra.
. It bears emphasis that what is herein described as “assessment of risk” is neither unprecedented nor unique to this area of law. To the contrary, assessment of risk is a normal part of judicial and administrative fact-finding. Thus EPA is not attempting to expand its powers; rather, petitioners seek to constrict the usual flexibility of the fact-finding process. Petitioners argue that the Administrator must decide that lead emissions “will endanger” the public health solely on “facts,” or, in the words of the division majority, by a “chain of scientific facts or reasoning leading [the Administrator] ineluctably to this conclusion * * Division op. at 59. Petitioners demand sole reliance on scientific facts, on evidence that reputable scientific techniques certify as certain. Typically, a scientist will not so certify evidence unless the probability of error, by standard statistical measurement, is less than 5%. That is, scientific fact is at least 95% certain.
Such certainty has never characterized the judicial or the administrative process. It may be that the “beyond a reasonable doubt” standard of criminal law demands 95% certainty. Cf. McGill v. United States, 121 U.S.App.D.C. 179, 185 n.6, 348 F.2d 791, 797 n.6 (1965). But the standard of ordinary civil litigation, a preponderance of the evidence, demands only 51% certainty. A jury may weigh conflicting evidence and certify as adjudicative (although not scientific) fact that which it believes is more likely than not. Since Reserve Mining was adjudicated in court, this standard applied to the court’s fact-finding. Inherently, such a standard is flexible; inherently, it allows the fact-finder to assess risks, to measure probabilities, to make subjective judgments. Nonetheless, the ultimate finding will be treated, at law, as fact and will be affirmed if based on substantial evidence, or, if made by a judge, not clearly erroneous.
The standard before administrative agencies is no less flexible. Agencies are not limited to scientific fact, to 95% certainties. Rather, they have at least the same fact-finding powers as a jury, particularly when, as here, they are engaged in rule-making.
Looking to the future, and commanded by Congress to make policy, a rule-making agency necessarily deals less with “evidentiary” disputes than with normative conflicts, projections from imperfect data, experiments and simulations, educated predictions, differing assessments of possible risks, and the like.
Amoco Oil Co. v. EPA, supra note 2, 163 U.S.App.D.C. at 175, 501 F.2d at 735. An agency’s finding of fact differs from that of a jury or trial judge primarily in that it is accorded more deference by a reviewing court. See note 74 infra. Thus, as a matter of administrative law, the Administrator found as fact that lead emissions “will endanger” the public health. That in so doing he did not have to rely solely on proved scientific fact is inherent in the requirements of legal fact-finding. Petitioners’ assertions of the need to rely on “fact” confuse the two terminologies. We must deal with the terminology of law, not science. At law, unless the administrative or judicial task is peculiarly factual in nature, or Congress expressly commands a more rigorous finding, see 21 U.S.C. § 355(d); cf. pages 395-396 of 176 U.S.App.D.C., pages 23-24 of 541 F.2d supra, assessment of risks as herein described typifies both the administrative and the judicial fact-finding function, and is not the novel or unprecedented theory that petitioners contend.
. As far as the cumulative impact theory is concerned, the relevant portions of the two provisions are as follows:
[The Administrator may regulate any automobile emission that] causes or contributes to, or is likely to cause or to contribute to, air pollution which endangers the public health or welfare.
Section 202(a)(1), 42 U.S.C. § 1857f-l(a)(l) (emphasis added).
[The Administrator may regulate a fuel or fuel additive] if any emission products of such fuel or fuel additive will endanger the public health or welfare * * *.
Section 211(c)(1)(A), 42 U.S.C. § 1857f-6c(c)(l)(A) (emphasis added).
. Since automobile emissions generate approximately 90% of all airborne lead, see'note 7 supra and accompanying text, the danger posed by lead-polluted air is virtually identical to that posed by lead automobile emissions. As the text suggests, the non-automotive sources of lead with which the cumulative impact theory is concerned are not airborne.
. Ethyl actually quotes Sen. Baker’s second use of the phrase, a few sentences after the one quoted in the text, see 116 Cong.Rec. 32920 (1970). It, however, is in the same context of contrasting permissible EPA action under § 211(c)(1)(A) and (B), and thus provides no more support for Ethyl’s position. The implausibility of Ethyl’s reading of Sen. Baker’s words is heightened by the fact that the Senator himself realized the important contribution of dietary lead to the total body lead burden and so informed the Senate only sentences before the quoted language. Id.
. While the incremental effect of lead emissions on the total body lead burden is of no practical value in determining whether health is endangered, it is of value, of course, in deciding whether the lead exposure problem can fruitfully be attacked through control of lead additives. Moreover, even under the cumulative impact theory emissions must make more than a minimal contribution to total exposure in order to justify regulation under § 211(c)(1)(A). We accept the Administrator’s determination that the contribution must be “significant” before regulation is proper. See 38 Fed.Reg. 33734.
. Congress had before it a complete explanation of the multiple sources of human lead exposure. It understood that lead is ubiquitous in nature, that trace elements of lead are present in everyone, and that only when lead concentration reaches higher levels would the public be endangered. It could not have thought that lead automobile emissions could, by themselves, endanger the public, although it clearly did think they could be regulated only if they provided a significant increment to the total human lead burden. See, e. g., Hearings on S. 3229, S. 2466 & S. 3546 before the Subcommittee on Air & Water Pollution of the Senate Committee on Public Works, 91st Cong., 2d Sess., pt. 1, at 433-434 (1970) (answers to Sen. Muskie’s questions, supplied by the Dept, of Health, Education & Welfare); id, pt. 3, at 1177; 116 Cong.Rec. 32920 (1970) (remarks of Sen. Baker).
. Considerable deference is owed to the interpretation of a statute by the officer charged with its administration. Train v. Natural Resources Defense Council, Inc., 421 U.S. 60, 87, 95 S.Ct. 1470, 1485, 43 L.Ed.2d 731, 750 (1975) (Clean Air Act); Zuber v. Allen, 396 U.S. 168, 192, 90 S.Ct. 314, 327, 24 L.Ed.2d 345, 359 (1969); Udall v. Tallman, 380 U.S. 1, 16, 85 S.Ct. 792, 801, 13 L.Ed.2d 616, 625 (1965); United States v. American Trucking Assns, 310 U.S. 534, 549, 60 S.Ct. 1059, 1067, 84 L.Ed. 1345, 1354 (1940). This deference is heightened when, as here, the interpretation is of a new statute by its implementing agency. Power Reactor Development Co. v. International Union of Electricians, 367 U.S. 396, 408, 81 S.Ct. 1529, 1535, 6 L.Ed.2d 924, 932 (1961); United States v. Zucca, 351 U.S. 91, 96, 76 S.Ct. 671, 674, 100 L.Ed. 964, 970 (1956); United States v. American Trucking Assns, supra; Norwegian Nitrogen Products Co. v. United States, 288 U.S. 294, 315, 53 S.Ct. 350, 358, 77 L.Ed. 796, 807 (1933); Natural Resources Defense Council, Inc. v. Train, 166 U.S.App.D.C. 312, 326, 510 F.2d 692, 706 (1975) (Clean Air Act). See note 16 supra.
.See, e. g., Parke v. Bradley, 204 Ala. 455, 86 So. 28 (1920); Forbes v. Board of Health, 28 Fla. 26, 9 So. 862 (1891); People ex rel. Bar-more v. Robertson, 302 Ill. 422, 134 N.E. 815, 22 A.L.R. 835 (1922); Blue v. Beach, 155 Ind. 121, 56 N.E. 89 (1900); Walker v. Sears, 245 Iowa 262, 61 N.W.2d 729 (1953); State ex rel. Anderson v. Fadely, 180 Kan. 652, 308 P.2d 537, 548 (1957); Board of Health v. Kollman, 156 Ky. 351, 160 S.W. 1052 (1913); Rock v. Carney, 216 Mich. 280, 185 N.W. 798, 22 A.L.R. 1178 (1921); State ex rel. Freeman v. Zimmerman, 86 Minn. 353, 90 N.W. 783 (1902); Crayton v. Larabee, 220 N.Y. 493, 116 N.E. 355 (1917); Salt Lake City v. Howe, 37 Utah 170, 106 P. 705 (1910).
. The statute demands “consideration” not only of the relevant scientific and medical evidence, but also of the possibility of regulation under § 202. Section 211(c)(2)(A), 42 U.S.C. § 1857f-6c(c)(2)(A). That means, of course, no more than it says: actual good faith consideration of the specified evidence and options, as reflected in the basis and purpose statement required by § 4(b) of the APA, 5 U.S.C. § 553(c). NPRA argues that the statute demands mandatory deference to the possibility of regulation under § 202 rather than mere consideration of such a possibility. NPRA Supp.Br. at 33-35. We reject the argument. NPRA ignores the deletion of the specific findings requirement from the original bill. As pointed out above, see pages 393-395 of 176 U.S.App.D.C., pages 21-23 of 541 F.2d, & note 14 supra, one effect of the deletion was to eliminate mandatory deference to § 202 and substitute instead “consideration” of § 202 regulation. While Congress thus indicated its preference for regulation under § 202, it left the Administrator with full discretion to implement that preference or not. See note 14 supra.
NPRA, joined by Nalco, further argues that EPA’s consideration of § 202 regulation was arbitrary and capricious because the agency rejected reliance on lead traps as a means of emission control. NPRA Supp.Br. at 35-37; Nalco Supp.Br. at 57-58. EPA did reject lead traps, but that rejection was not arbitrary and capricious. Rather, they were rejected because § 202 authorizes EPA to order emission controls only on new vehicles, which would largely be using lead-free gasoline anyway because of the catalytic converter. On new vehicles there would be no lead emissions to trap. See note 2 supra. The regulations before us seek to reduce lead emissions from in-use vehicles; for such a purpose § 202 is of no use. 38 Fed.Reg. 33737.
We find that EPA’s consideration of the evidence presented and of the alternatives to regulation under § 211 meets the good faith test suggested above. See Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136, 153 (1971). See also note 68 infra.
. This distinction is not unduly technical, since § 211(c)(1) expressly allows the Administrator to “control or prohibit” while § 211(c)(2)(C), drafted contemporaneously, repeatedly indicates that a finding is required only before a fuel or fuel additive is “prohibited.” Parallel § 211(c)(2)(A) and (B), on the other hand, also expressly apply before a fuel or fuel additive may be “controlled or prohibited.”
Nonetheless, the purpose behind § 211(c)(2)(C) — avoidance of counterproductive results and protection of the public health — suggests that a finding is of equal import whether the fuel or fuel additive is to be controlled or prohibited, so we do not conclude definitely that no such finding is required for control under § 211. Since the Administrator has furnished a finding such as is required by § 211(c)(2)(C), 38 Fed.Reg. 33737-33739 (1973), and we think that finding is adequate under the Amoco standards suggested below, we do not find it necessary to determine whether the finding was required in this case. See note 68 infra.
. Nalco argues that the Administrator’s recent decision to suspend the 1977 statutory emission standards for hydrocarbons and carbon-monoxide, 40 Fed.Reg. 11900, see note 2 supra, invalidates his assessment of the impact of the substitute additive, aromatic hydrocarbons. Nalco Supp.Br. at 45-50. Cf. NPRA Supp.Br. at 48-49. This argument appears to be a challenge to the regulations based on new information. Such challenges are cognizable under § 307 of the Clean Air Act, 42 U.S.C. § 1857h-5(b)(1), but may be brought to this court only after a preliminary presentment of the new information and a request for action to the agency. Oljato Chapter of Navajo Tribe v. Train, 169 U.S.App.D.C. 195, 207-208, 515 F.2d 654, 666-667 (1975). Since Nalco has not complied with the rule of Oljato Chapter, this new information claim does not appear to be properly before us.
In any case, however, Nalco’s claim is without merit. Undoubtedly high octane aromatic hydrocarbons will be used to compensate for the reduction in lead caused by the regulations and the average aromatic content of gasoline will rise from 22% to 29% (this increase includes that attributable to the lead-free regulations at issue in Amoco). Aromatics can result in emissions of polynuclear aromatic (PNA) hydrocarbons, some of which are carcinogenic.
Nonetheless, the Administrator determined that substitution of PNA emissions for lead emissions would be less dangerous to the public health because: (1) PNA emissions from automobiles, together with emissions from refineries that produce gasoline, account for only approximately 2% of PNA emissions in the ambient air; (2) PNA emissions are being reduced from automobiles by the applicable hydrocarbon standards; and (3) PNA automobile emissions will continue to decrease overall, even while the hydrocarbon content of gasoline increases, as uncontrolled automobiles are retired and replaced by those with emission control systems. 38 Fed.Reg. 33738. Thus the Administrator concluded that replacement of lead additives by increased hydrocarbons would result only in a slight slowing of the rate of decrease of PNA emissions from automobiles. We think this is a sufficient finding under § 211(c)(2)(C).
Nalco’s charge that suspension of emission standards changes this calculation is untrue. Even under the suspended standards, hydrocarbons will be significantly controlled and PNA emissions will continue to decrease. EPA calculated the effect of an increase in hydrocarbon content of gasoline under several anticipated circumstances, including the assumption that the interim standards would remain in effect through 1980. In that circumstance, which may turn out to be the correct forecast, PNA emissions would still decrease 69% over 1973 levels by 1980. Without the low-lead regulations they would decrease 72%, only a 3% better rate. JA 1428-1431; see also id. 1489-1490. Thus EPA anticipated the possibility of suspended emission standards, calculated the effects of that suspension on PNA emissions, and found them to be minimal. We cannot find this substitute additive finding to be insufficient.
. The hybrid procedures that placed the EPA action somewhere between informal rule-making and adjudication and that caused such extended inquiry by the Amoco court are largely absent from this case. In Amoco, in addition to the “findings” requirement of § 211(c)(2)(B), the court had to deal with the same section’s requirement of a public hearing. The resultant standard of review was, nonetheless, not significantly different from traditional “arbitrary and capricious” review. See Amoco Oil Co. v. EPA, supra note 2, 163 U.S.App.D.C. at 171-*406181, 501 F.2d at 731-741. In this case there is no public hearing requirement and the only “findings” requirement is that of § 211(c)(2)(C), which we have construed at pp. 404-405 of 176 U.S.App.D.C., at pp. 32-33 of 541 F.2d supra.
The requirement that the Administrator “consider” various evidence before acting under § 211(c)(1)(A) guides the direction, but does not affect the rigor, of “arbitrary and capricious” review. Such guidelines are typical as a prerequisite for agency action, see e. g., § 1002(e) (the Rule of Ratemaking) of the Federal Aviation Act of 1958, 49 U.S.C. § 1482(e)(1970), and assurance of agency compliance is simply one part of “arbitrary and capricious” review. Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136, 153 (1971) (court must determine “whether the decision was based on a consideration of the relevant factors”).
. All the Clean Air Act says about judicial review of EPA action under § 211 is that it shall be available exclusively in this court. 42 U.S.C. § 1857h-5(b)(1).
. Our review of the Administrator’s construction of the statute, see pages 383-406 of 176 U.S.App.D.C., pages 11-34 of 541 F.2d supra, is authorized by § 10(e)(2)(C), 5 U.S.C. § 706(2)(C), which allows the reviewing court to set aside agency action it finds to be “in excess of statutory jurisdiction, authority, or limitations, or short of statutory right * Our review of EPA procedures, see pages 420-425 of 176 U.S.App.D.C., pages 48-53 of 541 F.2d infra, is pursuant to § 10(e)(2)(D), 5 U.S.C. § 706(2)(D), which authorizes reversal of agency action taken “without observance of procedure required by law.”
. NPRA seeks to attach meaning to the fact that an express presumption of validity was included in the Senate version of the Clean Air Act, and then deleted by the conference committee. NPRA Supp.Br. at 21-24. See S.Rep. No. 91-1196, 91st Cong., 2d Sess., at 41, 125 (1970). As the above cited cases make clear, however, a general presumption of validity attaches to any regulation within an agency’s delegated powers. Thus the deletion is of no consequence.
. Of course, that basis must be expressed by the agency itself and not supplied by the court. SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 1577, 91 L.Ed. 1995, 1999 (1947). Nonetheless, a decision of “less than ideal clarity” will be upheld if the agency’s rationale “may reasonably be discerned.” Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 286, 95 S.Ct. 438, 442, 42 L.Ed.2d 447, 456 (1974). See also Colorado Interstate Gas Co. v. FPC, 324 U.S. 581, 595, 65 S.Ct. 829, 836, 89 L.Ed. 1206, 1219 (1945).
. Overton Park also requires the reviewing court to consider “whether there has been a clear error of judgment.” Citizens to Preserve Overton Park v. Volpe, supra note 69, 401 U.S. at 416, 91 S.Ct. at 824, 28 L.Ed.2d at 153, citing L. Jaffe, Judicial Control of Administrative Action 182 (1965); McBee v. Bomar, 296 F.2d 235, 237 (6th Cir. 1961); In re Josephson, 218 F.2d 174, 182 (1st Cir. 1954); Western Addition Community Organization v. Weaver, 294 F.Supp. 433 (N.D.Cal.1968); Wong Wing Hang v. INS, 360 F.2d 715, 719 (2d Cir. 1966). While as used, carefully bracketed by traditional statements of the restraint of “arbitrary and *407capricious” review, the phrase works no change in the law, the Court’s choice of language is troublesome. The phrase sounds much like the “clearly erroneous” standard used to review the factual findings of a trial court sitting without a jury. Rule 52(a), Fed.R. Civ.P. Unlike an agency determination or a jury verdict, such findings may be fairly readily reversed. District of Columbia v. Pace, 320 U.S. 698, 702, 64 S.Ct. 406, 408, 88 L.Ed. 408, 410 (1944); 4 K. Davis, supra note 15, § 29.02, at 118-126; L. Jaffe, supra, at 615-616. See especially Orvis v. Higgins, 180 F.2d 537, 540 (2d Cir.) (Frank, J.), cert. denied, 340 U.S. 810, 71 S.Ct. 37, 95 L.Ed. 595 (1950). Indeed, under “clearly erroneous” review a court may substitute its judgment for that of the trial court and upset findings that are not unreasonable. See 4 K. Davis, supra note 15, at 121-122.
Since Overton Park expressly forbade such intrusive review, 401 U.S. at 416, 91 S.Ct. at 823, 28 L.Ed.2d at 153, it plainly did not intend to use the “clear error of judgment” phrase to replace sub silentio “arbitrary and capricious” review with “clearly erroneous” review. Nonetheless, more than linguistic echoes of “clearly erroneous” review accompany the Court’s turn of phrase. To the extent the cases relied upon by the Court support consideration of “clear errors of judgment,” they all involve review of trial courts’, and not agencies’, abuses of discretion. See cases cited supra, and compare L. Jaffe, supra, at 182 with id. at 615-616. Such review is intrusive and essentially identical with “clearly erroneous” review. See McBee v. Bomar, supra; In re Josephson, supra. On the other hand, the Court also cited a case in which Judge Friendly recognized that an agency’s abuse of discretion, unlike a court’s, is reviewed under the “arbitrary and capricious” standard, 5 U.S.C. § 706(2)(A), and that review in such cases should be much more deferential than under the “clearly erroneous” standard. Wong Wing Hang v. INS, supra, 360 F.2d at 718-719.
All of this makes the Court’s intent in Over-ton Park somewhat difficult to plumb and its standard even more uncertain of application. We do not think the Court’s use of the “clear error of judgment” phrase was an attempt vastly to revamp traditional “arbitrary and capricious” review. See infra. Nonetheless, we fear its use of this phrase so familiar to judges in another, and significantly more intrusive, context may unintentionally prompt judicial distortion of the “arbitrary and capricious” standard. Already at least one court has expressly indicated that it is prepared to read the Court’s use of the phrase as approval of intrusive “clearly erroneous” review of agency action. Raitport v. National Bureau of Standards, 385 F.Supp. 1221, 1225 (E.D.Pa.1974). Meanwhile, other courts use the “clear error of judgment” phrase as a shorthand summary of “arbitrary and capricious” review, see, e. g., Union Electric Co. v. EPA, 515 F.2d 206, 216 (8th Cir.), cert. granted, 423 U.S. 821, 96 S.Ct. 35, 46 L.Ed.2d 38, 44 U.S.L. Week 3200 (1975); Conservation Council of North Carolina v. Froehlke, 473 F.2d 664, 665 (4th Cir. 1973); Schicke v. United States, 346 F.Supp. 417, 420, 422-423 (D.Conn.1972), while others loosely treat the phrase as synonymous with “clearly erroneous.” Duke City Lumber Co. v. Butz, 382 F.Supp. 362, 374 (D.D.C.1974) (“whether the agency’s threshold decision was arbitrary, capricious or otherwise clearly erroneous”); Daly v. Volpe, 350 F.Supp. 252, 255 (W.D.Wash.1972) (“it was not clearly erroneous, and therefore it was not arbitrary and capricious”).
Post-Overton Park decisions, as well as the internal evidence in Overton Park itself, see supra, have made clear that the Court does not intend the “clear error of judgment” phrase to sanction review more intrusive than traditional “arbitrary and capricious” review; rather, the Court has reaffirmed that the reviewing court must defer if the agency has a rational basis for its decision. Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra note 73, 419 U.S. at 290, 95 S.Ct. at 444, 42 L.Ed.2d at 458; United States v. Allegheny-Ludlum Steel Corp., 406 U.S. 742, 749, 92 S.Ct. 1941, 1946, 32 L.Ed.2d 453, 460 (1972). See note 79 infra. Thus it is important that courts not think themselves licensed to embark upon wide-ranging searches for “clear errors of judgment.” Such searches can only distort the established appellate role in reviewing informal agency action. Rather, we think Overton Park’s troublesome phrase is best read as no more than an affirmation of the traditional standard of review. Accordingly, in the context of “arbitrary and capricious” review, we shall reverse for a “clear error of judgment” only if the error is so clear as to deprive the agency’s decision of a rational basis.
. While Greater Boston Television Corp. v. FCC, 143 U.S.App.D.C. 383, 392, 444 F.2d 841, 850 (1970), cert. denied, 403 U.S. 923, 91 S.Ct. 2229, 2233, 29 L.Ed.2d 701 (1971), was a substantial evidence case, its statement of the proper scope of a reviewing court’s inquiry into the evidence has been adopted by the informal rule-making cases cited above.
. This rule has been most directly stated in cases involving substantial evidence review, see, e. g., Consolo v. Federal Maritime Commission, 383 U.S. 607, 620, 86 S.Ct. 1018, 1026, 16 L.Ed.2d 131, 140 (1966); Universal Camera Corp. v. NLRB, 340 U.S. 474, 488, 71 S.Ct. 456, 464, 95 L.Ed. 456, 467 (1951); Environmental Defense Fund, Inc. v. ERA (Coahoma), supra note 28, 160 U.S.App.D.C. at 130, 489 F.2d at 1254, but it is nonetheless applicable here as well. The requirement of affirming decisions with which we disagree is inherent in Overton Park’s command that “[t]he court is not empowered to substitute its judgment for that of the agency.” Citizens to Preserve Overton Park v. Volpe, supra note 69, 401 U.S. at 416, 91 S.Ct. at 824, 28 L.Ed.2d at 153.
. In a substantial evidence case Judge Wilkey has well delineated our limited role:
*409In the case at bar our task is made somewhat simpler than the agency’s by adhering conscientiously to the proper scope of judicial review of administrative action, i. e., we as a court are confronted with a problem in administrative law, not in chemistry, biology, medicine, or ecology. It is the administrative agency which has been called upon to hear and evaluate testimony in all scientific fields relevant to its ultimate question of permission or prohibition of the sale and use of DDT. The EPA Administrator had an opportunity to make a careful study of the record of seven months of public hearings and the summaries of evidence prepared for him, heard oral argument, and now has arrived at a decision to ban most uses of DDT. It is his decision which we must review; we are not to make the same decision ourselves.
Environmental Defense Fund, Inc. v. EPA (Coahoma), supra note 28, 160 U.S.App.D.C. at 128, 489 F.2d at 1252 (emphasis in original). See also note 53 supra.
. In stating the scope of review of informal rule-making by the Interstate Commerce Commission, the Supreme Court succinctly summarized these principles:
We do not weigh the evidence introduced before the Commission; we do not inquire into the wisdom of the regulations that the Commission promulgates, and we inquire into the soundness of the reasoning by which the Commission reaches its conclusions only to ascertain that the latter are rationally supported.
United States v. Allegheny-Ludlum Steel Corp., supra note 74, 406 U.S. at 749, 92 S.Ct. at 1946, 32 L.Ed.2d at 460.
. Review for substantial evidence is mandated only for agency adjudications and formal rule-making proceedings. 5 U.S.C. § 706(2)(E). Since “arbitrary and capricious” review does not involve determining whether the agency decision is supported by substantial evidence, it is considered the more lenient form of review. Nonetheless, some have noted that in reviewing the evidence relied upon in agency proceedings, the two standards often seem to merge. Associated Industries of New York State, Inc. v. United States Department of Labor, 487 F.2d 342, 349-350 (2d Cir. 1973) (Friendly, X). The primary difference between the two in such cases would seem to be that “substantial evidence” review is limited to evidence developed in formal hearings, while “arbitrary and capricious” review of an agency engaged in informal rule-making is not so limited, but rather may consider the agency’s developed expertise and any evidence referenced by the agency or otherwise placed in the record. Market St. Ry. v. Railroad Comm’n, 324 U.S. 548, 559-561, 65 S.Ct. 770, 776-777, 89 L.Ed. 1171, 1180-1182 (1945); City of Chicago v. FPC, 147 U.S.App.D.C. 312, 322-326, 458 F.2d 731, 741-745 (1971), cert. denied, 405 U.S. 1074, 92 S.Ct. 1495, 31 L.Ed.2d 808 (1972).
.As Chief Justice Shaw observed over a century ago, “inferences drawn from independent sources, different from each other, but tending to the same conclusion, not only support each other, but do so with an increased weight.” Commonwealth v. Webster, 59 Mass. (5 Cush.) 295, 317 (1850).
.Both petitioners and the Wilkey dissent (hereinafter dissent) engage in a detailed attack on the evidentiary basis for the Administrator’s decision. Their arguments are considered in corresponding detail in the appendices to this opinion.
.See note 60 supra.
. Indeed, at the conclusion of Section IV of the Third Health Document, “Can an Acceptable Lead Body Burden be Defined?,” the Agency suggested seven important areas for future inquiry. Third Health Document at IV-7, 8, JA 75-76. Ethyl claims these questions only demonstrate “that EPA is unable at this time to support its blood lead level theories with pertinent data * * Ethyl brief at 38. To the contrary, the questions only show the candor with which the Agency has approached this important issue. Sufficient data exists for the Administrator to assess risks and make a judgment within the “will endanger” standard. That further data would enable the Agency to deal more precisely with the problem does not undercut the Administrator’s judgment. It is the nature of scientific inquiry that further data will always permit such refinement.
. Petitioners Ethyl and Nalco specially attack the Administrator’s reliance on these subclinical effects of exposure to low lead levels both here and in assessing the damages from dust-fall. Ethyl Supp.Br. at 45; Nalco Supp.Br. at 30-34. In fact, however, petitioners merely repeat the very qualifications on the evidence that led the Administrator to discount its value. Petitioners and the Administrator agree far more than they disagree on this issue. We cannot say it was irrational for the Administrator to note the possibility of subclinical effects, nor can we say it was irrational for him largely to discount their probability. Accordingly, we must reject petitioners’ argument.
.A few examples indicate the kind of evidence before the Administrator:
A study by David et al. associating increased frequency of hyperactivity among children with blood lead levels of 25-55 ug. Cited in Third Health Document at III-6, 7, JA 59-60.
A study by Piomelli et al. which found that 55% of children tested with blood lead levels of 40-59 ug and 100% of those tested with blood lead levels in excess of 60 ug had demonstrable evidence of metabolic interference with heme synthesis in the bone marrow. Cited in Third Health Document at IV-3, 4, 5, JA 71-73.
*412A study by Tola et al. finding mild anemia in lead workers with blood lead levels in the 40-60 ug range. Cited in Third Health Document at III — 6, JA 59.
A study by Pueschel et al. suggesting that in children lead body burdens below those usually associated with clinical lead poisoning may contribute to renal as well as neurological damage. Cited in Third Health Document at III-5, JA 58.
. Tepper & Levin, “A Survey of Air and Population Lead Levels in Selected American Communities,” Dept, of Environmental Health, College of Medicine, U. Cincinnati, Ohio (EPA Contract PN 22-68-28) (Dec. 1972), JA 840 (hereinafter Seven Cities Study).
. Ethyl also once again attacks the mere fact that the Administrator exercised his judgment. See Ethyl brief at 47.
. Petitioners PPG Industries, Inc. (PPG) and E. I. duPont de Nemours & Company (duPont) charge that this conclusion has no bearing on whether there is undue lead exposure among the general adult population. PPG/duPont Supp.Br. at 19. Of course it does not. It does have bearing, however, on whether blood lead levels are elevated among children and among the general population, both adult and child. It is for this purpose that the conclusion is cited.
Ethyl argues that these children have high blood lead levels primarily because of exposure to lead-based paint. Ethyl Supp.Br. at 40-41. The Administrator does not disagree. 38 Fed. Reg. 33735. What is important for his purpose is that blood lead levels are elevated in the general public; under the cumulative impact theory the source of the elevation is irrelevant in determining whether the public health is endangered by exposure to lead. See pages 401-403 of 176 U.S.App.D.C., pages 29-31 of 541 F.2d supra. The source becomes relevant only when deciding whether the emission products of lead additives make a significant contribution to that exposure, thereby justifying regulation under § 211(c)(1)(A).
.NPRA, PPG and duPont suggest that if certain occupational groups are overexposed to lead concentrations in the air, they should be protected under the Occupational Safety and Health Act of 1970, 29 U.S.C. § 651 et seq., rather than the Clean Air Act. NPRA Supp.Br. at 6-8; PPG/duPont Supp.Br. at 20. Other than by issuing mailmen, policemen, and service station attendants gas masks, we are baffled as to how special occupational regulations could protect such overexposed groups, who are overexposed only because they must work outdoors on or near busy city streets. We have no doubt, however, that the Clean Air Act was intended to protect such people. As Sen. Muskie stated “for the purposes of the record”:
In setting standards, not only ambient standards but emission standards, * * * we must be concerned with the health effect upon the most vulnerable in our population
Hearings on S. 3229, S. 3455, S. 3546 before the Subcomm. on Air and Water Pollution of the Senate Comm, on Public Works, 91st Cong., 2d Sess, at 74 (1970). No less than with “particu*413larly sensitive citizens such as bronchial asthmatics and emphysematics” who receive the special attention of the Clean Air Act, S.Rep. No.91-1196, 91st Cong., 2d Sess. 10 (1970), citizens who are occupationally exposed outdoors to otherwise uncontrollable lead concentrations in the ambient air are among “the most vulnerable in our population” and deserve to be considered when regulations under § 211(c)(1)(A) are at issue.
. The dissent criticizes this comparison of the Administrator’s use of studies of occupational groups with the Eighth Circuit’s use of similar evidence on the ground that the court explained the validity of its use of occupational data whereas the Administrator did not. Dissent at 476 of 176 U.S.App.D.C., at 104 of 541 F.2d. The Administrator, however, based his inferences on studies of occupational groups whose lead intake results from exposure, albeit for longer than average periods of time, to the same air breathed by the general public. See pp. 412-413 of 176 U.S.App.D.C., pp. 40-41 of 541 F.2d supra. The Eighth Circuit, by contrast, was explaining the validity of inferring a hazard to the general public from studies showing a hazard to workers and others exposed to higher levels of a somewhat different substance than that with which the court was concerned. See 514 F.2d at 511-512. In short, the Eighth Circuit was faced with a situation where differences between the hazard facing the occupationally exposed group and the hazard facing the general public required that the propriety of reaching a conclusion about the public from occupational studies be explained; the Administrator was not.
. EPA specifically requested comment on the figures it used in calculating the respiratory absorption of lead by a “standard man.” 37 Fed.Reg. 11787 (1972), JA 21.
. The dissent suggests that this court, not the Administrator, first propounded this rationale for relying on the intrametropolitan results of the epidemiological studies. Dissent at 410 of 176 U.S.App.D.C., at 38 of 541 F.2d. This critique ignores the facts that the Administrator referred to this explanation at the beginning of his opinion, JA 3, and that the Third Health Document identifies the language in the Seven Cities Study which justifies the Administrator’s reliance on that study’s intrametropolitan area data. JA 89, quoting JA 844 (the introductory summary of the Seven Cities Study). The section of the Seven Cities Study summarized by the language quoted in the Third Health Document explicitly states the rationale relied on by the Administrator. JA 892; see pp. 429 to 430 of 176 U.S.App.D.C., pp. 57 to 58 of 541 F.2d, & note 4 infra. Clearly, the “agency’s path” to its decision to credit the intrametropolitan results “may reasonably be discerned.” Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra note 73, 419 U.S. at 286, 95 S.Ct. at 442, 42 L.Ed.2d at 456.
Despite its origins in the Seven Cities Study itself, the dissent rejects this explanation for the Administrator’s approach to the evidence as totally irrational. The dissent argues that it is illogical to maintain that intrametropolitan dietary lead can be assumed to be relatively constant despite the income differences in a large metropolitan area. This argument, which has no foundation in any material before this court, illustrates the danger of “plausible-sounding, but simplistic, judgments of the relative weight to be afforded various pieces of technical data.” Concurring op. of Chief Judge Bazelon at 438 of 176 U.S.App.D.C., at 66 of 541 F.2d. Without ourselves attempting to assume a scientific role, we note that since the main source of dietary lead is the soil, see note 5 supra, the geographical origin of foodstuffs *415may be a more important determinant of their lead content than is price. In any event, whatever the underlying basis for this approach might be, the important point is that we have no reason to fault the Administrator’s decision to rely on the scientific expertise of the authors of the epidemiological studies before him.
. Petitioners’ persistent refusal to recognize that these are the reasons the Administrator is concerned about lead dustfall makes their constant arguments about leaded paints completely unresponsive to the Administrator’s analysis of the evidence.
. Ter Haar & Aronow, New Information on Lead in Dirt and Dust as Related to the Childhood Lead Problem, presented at the EPANIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C., Oct. 1, 1973, JA 917, 922 (Fig. 1).
. Since proof of the danger posed by inhalation of asbestos wastes was much stronger than proof of the danger posed by ingestion the Reserve Mining court ordered immediate action taken with regard to the discharge of wastes into the air. See note 34 supra.
. The dissent states that its disagreement with the Administrator and the majority concerning the interpretation of the “will endanger” standard “reduces itself to semantics.” Dissent at 445 of 176 U.S.App.D.C., at 73 of 541 F.2d. The dissent also rejects the Administrator’s reliance on the dustfall hypothesis simply because it is a hypothesis. Dissent at 413-414 ’of 176 U.S.App.D.C., at 41-42 of~54l F.2d. Yet as we have demonstrated above and *419as the Reserve Mining court held, a reasonable hypothesis supported by evidence is a sufficient basis for regulating under the “will endanger” standard as the Administrator has interpreted that standard.
. This fact completely rebuts petitioners’ claim that this case does not present problems “on the frontiers of scientific knowledge.” Industrial Union Department, AFL-CIO v. Hodgson, supra note 35, 162 U.S.App.D.C. at 338, 499 F.2d at 474. See, e. g., NPRA Supp.Br. at 49. Although petitioners note that lead additives have been in use for over 50 years, and that questions about their safety have been raised for almost as long, see note 3 supra, the fact remains that virtually all the evidence cited by both EPA and petitioners was developed in the last five years, most of it for the purpose of supporting or opposing these regulations. Cf. note 3 supra. Moreover, both EPA and petitioners admit much more remains to be discovered. See pp. 397-398 of 176 U.S.App.D.C., pp. 25-26 of 541 F.2d and note 83 supra.
. The dissent’s repeated assertion that the evidence was insufficient to convince the independent scientific community, represented by other Government agencies, is unconvincing for two reasons. First, the Department of Health, Education, and Welfare, which shares with EPA responsibility for and scientific expertise in the areas of public and environmental health, endorsed the regulations. The dissent relies on a Jan. 29, 1973 letter from then Secretary of HEW Elliot Richardson to demonstrate the Department’s opposition to the lead reduction regulations. Dissent at 448-449 of 176 U.S.App.D.C., 76-77 of 541 F.2d. The Aug. 7, 1973 letter of Richardson’s successor, Caspar Weinberger, in which the Department endorsed removing lead from gasoline provided the result would not be the introduction of more harmful additives, JA 2507, is unconvincingly discounted. Dissent at 448-449 of 176 U.S.App.D.C., at 76-77 of 541 F.2d n.27. We have found reasonable the Administrator’s determination that reducing the lead content of gasoline would not have this perverse effect on the public health. See pp. 404-405 of 176 U.S.App.D.C., pp. 32-33 of 541 F.2d & notes 67-68 supra.
Second, the dissent’s equation of the views of other Government agencies with the views of the independent scientific community is itself fallacious. Without in the least impugning their motivation, it may be assumed that agencies such as the Department of the Interior and the Department of Commerce, see dissent at 446,448-449 of 176 U.S.App.D.C., at 74,76-77 of 541 F.2d respond to and represent interests other than those concerned with the environment and the public health. The very existence of EPA is evidence that Congress believed the other agencies of the Government could not adequately appraise and act against environmental threats. Truly independent scientific opinion, represented by comments from those with no association with environmental groups or industry, favored the regulations by a margin of approximately two to one. JA 5.
We also note that the California Air Resources Board has recently adopted regulations imposing a more stringent lead reduction schedule than is imposed by the regulations before us. See N.Y. Times, Feb. 20, 1976, at 12 col. 3.
. Since EPA never stopped accepting comments on the proposed regulations, the comment period on these regulations actually lasted almost two years. All comments were made public when they were received, in accordance with the procedure established by EPA’s notices of proposed rule-making. See, e. g., 38 Fed.Reg. at 1260, JA 17; 17 Fed.Reg. at 3882, JA 23. Moreover, in keeping with his statutory mandate to consider “all relevant medical and scientific evidence available to him,” § 211(c)(2)(A), 42 U.S.C. § 1857f-6c(c)(2)(A) (emphasis added), the Administrator considered comments received throughout this extended comment period. See p. 425 of 176 U.S.App.D.C., p. 53 of 541 F.2d & note 123 infra.
NPRA claims that placing the new studies in the public file did not satisfy EPA’s obligations because “EPA did not notify the public of the availability of these documents or that it intended to rely upon them in promulgating the final regulations.” NPRA Reply Br. at 17. This statement is not accurate, since EPA’s public notice of the proposed low-lead rule-making, supra, stated that all new information received concerning the proposed regulations would be made available for public inspection at its Office of Public Affairs. 38 Fed.Reg. at 1260, JA 17. Moreover, NPRA’s own submissions refer to documents contained in the EPA public file. See JA 1962; Doc. 1259. See also Doc. 1089 (letter from Ethyl, dated Oct. 12, 1973, requesting that certain documents not yet in the public file be placed there). Since the statute requires the Administrator to consider all evidence available to him, the complaint that petitioners could not know the Administrator would make use of the material in the public file is unimpressive.
. The reproposed regulations had made clear that these two alternatives were under consideration and that, at the time, the Administrator favored leaded pool averaging. 38 Fed.Reg. 1260 (1973). When the majority of industry comments favored use of the total pool system, 38 Fed.Reg. 33739 (1973), the Administrator shifted his position. While we think the notice at 38 Fed.Reg. 1260 was a sufficient basis for a change in position in the final regulations, the Administrator gave actual notice of the change to the refiners several months before the final regulations were promulgated. See NPRA letter dated Oct. 25, 1973, JA 1962. Moreover, petitioner NPRA submitted comments on that change. See id.
. Since all material information was made available for comment by petitioners and the public, it is not necessary for us to determine whether the Administrator may rely on information not previously made available.
. By contrast, the division majority allotted less than one and a half pages of its 73-page opinion to this argument. . Petitioners also treated this contention as a minor part of their attack on the regulations. The major thrust of the dissent’s complaint appears to be that although all of the evidence on which the Administrator relied was made available to the public at least three months before the regulations were issued, that evidence was not specifically delivered to or called to the attention of petitioners when it arrived at EPA. The dissent does not explain the origin of this supposed responsibility of the Agency. Certainly the EPA’s requirement of opportunity for notice . and comment is satisfied by placing evidence in a designated public access file and repeatedly announcing the existence and location of that file in the Federal Register. See note 99 supra. Thus the dissent’s chart of the availability of new evidence, dissent at 465 of 176 U.S.App.D.C., at 93 of 541 F.2d, is fundamentally misleading. All interested parties had legally adequate notice of the evidence relied on by the Administrator, and of his intent to rely, see note 99 supra, from the time that evidence was placed in the public file.
The dissent attempts to avoid this fact by repeatedly lamenting that EPA’s public file was “poorly indexed” (and dusty). Dissent at 451 n.34 454, 462-463 of 176 U.S.App.D.C., at 79 n.34, 82, 90-91 of 541 F.2d. On examination this entire attack turns out to be based on the excuse offered by EPA’s attorney for a slight delay in providing a document to this court. See id., at 451 n.34, of 176 U.S.App.D.C., at 79 n.34, of 541 F.2d. Had petitioners complained that they were not able to use the public file, we would of course have treated that complaint as a serious matter. But petitioners made no such allegation, nor did they suggest that EPA’s public file procedures were in any way less adequate than the procedures customarily followed by administrative agencies. It may well be that the public reference procedures of this still relatively new agency can be improved. But we see no basis for a hypothesis, really a speculation, that persons affected are so unable to use the public file that minimal fairness is lacking. In any event that complaint was not made by able and experienced counsel for petitioners.
. Rabinowitz et al., “Study of human lead metabolism using stable isotope tracers,” paper presented at EPA-NIEHS conference on Low Level Lead Toxicity, Raleigh, N. C., Oct. 1-2, 1973, JA 678; Rabinowitz et al, “Lead Metabolism in the Normal Human: Stable Isotope Studies,” 182 Science 725-727 (1973), JA 704.
. See also EPA letter of Oct. 2, 1975. That letter and a letter of Sept. 26, 1975 were responses to requests for information from this court. Copies of these letters were sent by EPA to counsel for all parties to this litigation. No party commented on either letter.
. The final draft was received by .EPA’s Office of the General Counsel Aug. 6, 1975, for transmittal to Ethyl, see note 117 infra, and to the public file. EPA letter of Sept. 26, 1975.
. Nalco Chemical Co. submitted extensive comments on the scientific evidence, including comments directed specifically at the lead isotope study, on Nov. 19, 1973. Doc. 821 at 7. Ethyl also submitted comments on this study in the form of a transcript of the October conference. Doc. 433.
The dissent finds it strange that we consider this transcript a substantive comment on the studies discussed at the meeting. Dissent at 461,462-463 of 176 U.S.App.D.C., at 68,90-91 of 541 F.2d. In response, we simply note that the only possible interpretation of Ethyl’s action in preparing and submitting the transcript was that the company wished to have the critical comments of its scientists and some others who attended the meeting brought to EPA’s attention as substantive critiques of the studies. A covering letter from Ethyl’s counsel which accompanied the company’s submission of the transcripts of a Feb. 26 and a March 15 meeting explicitly states this purpose. See JA 986.
. Tsuchiya ei al., Study of Lead Concentrations in Atmosphere and Population in Japan (undated), JA 1092.
. The study was received by EPA’s Office of the General Counsel on July 22, 1973, for transmittal to Ethyl, see note 117 infra, and to the public file. EPA letter of Sept. 26, 1975. It too was commented on by Nalco. Doc. 821 at 7. Since the study was available for comment four months prior to issuance of the Administrator’s decision, we find the significance of the fact that it was not cited in the Third Health Document, dissent at 452 of 176 U.S.App.D.C., at 80 of 541 F.2d, mystifying.
. Knelson et al., Kinetics of Respiratory Lead Intake in Humans, Proceedings of the International Symposium on Environmental Aspects of Lead 391-401 (1973), JA 596.
. Compare Doc. 85, JA 596-601, with Doc. Ill at 391-401. The final printed “version” differs from the “preprint” only in that the former contains a translation of the authors’ summary into French and German.
. Tepper & Levin, “A Survey of Air and Population Lead Levels in Selected American Communities,” Department of Environmental Health, College of Medicine, University of Cincinnati, Cincinnati, Ohio (1972), JA 840.
.The dissent’s emphasis on the importance of the reanalysis of the Seven Cities Study appears to be based on a misunderstanding of the meaning of statistical significance. Statistical significance simply expresses the level of assurance we can have that certain data was not the product of random relationships. See, e. g., F. Mosteller, R. Rourke & G. Thomas, Probability With Statistical Applications 304-307 (2d ed. 1970).
Unfortunately, the word “significant” in everyday use means not only “suggestive” but also “important” and “weighty”. We do not carry these everyday meanings over into statistics. * * * A statistically significant result is one that the data support as showing a real effect, as opposed to a result that might readily arise from sampling variation.
Id. at 307.
Typically, scientists refuse to certify an observed relationship as “significant” unless they are 95% certain that the data could not have been generated randomly. See note 58 supra. Thus the authors of the Seven Cities Study found the relationship between air and blood lead levels between cities “[n]ot significantly different from 0 at the 5% level,” JA 902, meaning that they could not be certain that if there were no genuine relationship in the real world, the relationship that was observed in this study would be found in at most five out of every 100 randomly drawn data samples. On reanalyzing the data using more refined techniques, EPA’s scientists determined that the observed relationship could be expected to be produced by chance less than one time out of 1,000. See Doc. 228 at Table 3. Similarly, the authors of the Seven Cities Study themselves recognized the implication of their “consistent observation” that blood lead levels were higher in a given metropolitan area’s urban region than in its suburbs. See p. 414 of 176 U.S.App.D.C., p. 42 of 541 F.2d, & note 92 supra. The reanalysis simply demonstrated that this “consistent observation” has less than a one in 1,000 likelihood of being the product of chance. See Doc. 228 at Table 3.
There is, of course, no reason why the Administrator cannot rely on observed relationships among data which he is less than 95% certain reflect a true underlying relationship between the phenomena that the data measure. See note 58 supra. The dissent’s assertion that only the reanalysis provides support for the Administrator’s decision assumes that the Administrator must be 95% certain before he credits any evidence.
. The purpose and contents of the reanalysis are described by its first sentence; “This brief paper is an attempt to answer questions raised by reviewers about the statistical analyses performed [in the Seven Cities Study].” Doc. 228 at 1.
. The results and methodology of the reanalysis were presented and discussed at a meeting of the EPA Hazardous Materials Advisory Committee on Feb. 26, 1973. JA 989-992. Representatives of Ethyl attended the meeting, prepared a transcript of it, including critical comments by some of those in attendance, and forwarded the transcript to the Administrator. JA 985; see note 106 supra. The dissent strives mightily to show that the reanalysis was not in the public domain until four days after the close of the formal comment period, rather than 13 days before. Dissent at 452-453 of 176 U.S.App.D.C., at 80-81 of 541 F.2d. - Even if this were true, the important point is that the reanalysis was available for public comment ■ some nine months prior to issuance of the regulations.
. Margulis et al., Residential Location, Ambient Air Lead Pollution and Childhood Lead Poisoning, — Arch Env. Health — (in press at time record was compiled), JA 626; Sayre et al., House and Hand Dust as a Potential Source of Childhood Lead Exposure, — Am.J.Dis.Child — (in press at time record was compiled); and Vostal et al., “Lead Containing House Dust: Another Source of Increased Lead Exposure in Inner City Children,” paper presented at EPANIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C., Oct. 1-2, 1973, JA 720, 738; Needleman & Shapiro, “Dentine Lead Levels in Asymptomatic Philadelphia School Children: Subclinical Exposure in High and Low Risk *424Groups,” paper presented at EPA-NIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C., Oct. 1-2, 1973, JA 662; Sachs, “Effects of a Screening Program on Changing Patterns of Lead Poisoning,” paper presented at EPANIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C., Oct. 1-2, 1973, JA 718.
. Compare JA 1-5 (the text of the Administrator’s decision) with JA 186-193 and JA 231-239 (discussions of the dustfall hypothesis in the Second Health Document).
. The dissent cites no authority at all in that part of its argument which focuses on the asserted need for notice of intent to rely. See dissent at 456-457, 461-466 of 176 U.S.App.D.C., at 84-85, 89-94 of 541 F.2d. The authorities cited in the same section of the dissent, see id. at 450-452, and nn.78-80 of 176 U.S.App.D.C., at 85-86 and nn. 78-80 of 541 F.2d, support only the proposition that an agency may not keep secret the purpose of its proceeding or information important to its decision. The citation to K. Davis, Administrative Law Treatise § 15.10 at 402 (1958), dissent at 458 n.78 of 176 U.S.App.D.C., at 86 n.78 of 541 F.2d, refers to a discussion of the “Procedures for Using and Challenging Extra-Record Facts” of which an administrative decision-maker has taken notice during an adjudicatory proceeding. Since all of the studies at issue here were disclosed by the Administrator as soon as they became available to him, not kept secret, and since there could be no allegation that any of this material was not part of the record properly before the Administrator, these authorities are of no assistance to the dissent’s argument.
Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., supra note 73, quoted, dissent at 458 of 176 U.S.App.D.C., at 86 of 541 F.2d also offers no support for the proposition that the APA requires notice of what material received and made public after publication of a proposed rule will be relied on to support the final regulation. In the first place, Bowman was an adjudicatory proceeding, not a notice-and-comment rule-making. More importantly, in the sentence immediately following the language quoted by the dissent the Supreme Court noted that a party’s right “to be apprised of the factual material on which the agency relies for decision” does “not preclude a factfinder from observing strengths and weaknesses in the evidence that no party identified.” 419 U.S. at 288 n.4, 95 S.Ct. at 443 n.4, 42 L.Ed.2d at 457 n.4. The Administrator’s treatment of the Seven Cities Study, as well as of the Japanese study submitted by lead industry opponents of the proposed regulation, fits *425squarely within this observation. Moreover, the Bowman Court also noted that the petitioners there “are not in a position to claim unfair surprise” because the Commission had offered the same rationale for its interpretation of evidence in a case decided at the time the hearings in Bowman were beginning. Id. at 289 n.4, 95 S.Ct. at 443 n.4, 42 L.Ed.2d at 457 n.4. That rationale, it should be noted, was potentially rebuttable. In the face of this conclusion, the assertion that petitioners were not on notice that the agency might rely on information in the public record of this rule-making proceeding appears frivolous.
Finally, we note that nothing in Wright, The Courts and the Rulemaking Process: The Limits of Judicial Review, 59 Cornell L.Rev. 375 (1974), supports the dissent’s attempt to transmute the three-step process established by § 4 into a potentially unending and fruitless series of notices, comments, and notices of intent to rely on comments.
. See note 99 supra. Indeed, petitioner Ethyl was directly furnished with all such documents as a result of a Freedom of Information Act, 5 U.S.C. § 552 (1970), suit. JA 1509-1510. See Ethyl Corp. v. EPA, 478 F.2d 47 (4th Cir. 1973).
. The dissent’s effort to demonstrate that little comment was received about the most recent of the studies relied on by the Administrator is irrelevant. All the APA requires is that there have been an opportunity to comment. The three months between the time the important new studies were placed in the public file and the time of the Administrator’s decision is clearly sufficient opportunity.
. JA 1474; JA 1502.
. Doc. 141. Chapters I, II, V, VII, and VIII (the conclusions) of this draft, dated Oct. 19, 1973, differ only in minor wording changes from the final Third Health Document. Chapters III, IV, and VI of the final version contain a few additional inconsequential paragraphs and references. The cover of the Oct. 19 draft states that “[i]t is being circulated for comment on its technical accuracy and policy implications.”
The early October preparation and circulation of a draft decision and a draft Third Health Document, both in obviously near final form, belie the dissents’ contention that this court’s 30-day order, see p. 382 of 176 U.S.App.D.C., p. 10 of 541 F.2d supra, forced EPA to rush into a decision it was not yet ready to make.
. See e. g., JA 2577 (Treasury Department comments); JA 2651 (letter from Secretary of Transportation); JA .2493 (Department of the Interior comments, dated Dec. 7, 1973). The Nalco comments, see note 106 supra, were addressed to both the draft Third Health Document and the draft of the Administrator’s decision.
. See Ethyl letter dated Oct. 29, 1973, Doc. 1090, presenting evidence considered in the Third Health Document, JA 79 (ref. 11) & 140 (ref. 43). The Ethyl letter called this evidence to EPA’s attention because “[t]o our knowledge these highly significant articles have not been referred to by EPA to date in its consideration of the health question * * Doc. 1090 supra. The timing of this letter suggests that Ethyl was referring to EPA’s failure to reference these articles in the Oct. 19 draft of the Third Health Document. Ethyl would have received that draft due to its Freedom of Information Act suit. See note 118 supra.
. Beyond their attacks on the Administrator’s interpretation of the statutory language, the evidence, and the procedures, petitioners raise a host of other objections to the final EPA regulations. We have considered above petitioners’ claims that the Administrator did not properly consider the possibility of regulation under § 202, see note 66 supra, and that he did not make a valid finding that any substitute additive would be less dangerous than lead, particularly in light of his recent suspension of 1977 statutory emission standards. See notes 67-68, supra. Cf. note 2 supra. We shall briefly address petitioners’ remaining contentions.
Ethyl claims a right to cross-examine agency witnesses. Ethyl Br. at 55. Not only does the *426Clean Air Act not require any hearings at all, but even if hearings were required cross-examination is not mandated in § 4 type proceedings, 5 U.S.C. § 553. United States v. Florida East Coast R. Co., 410 U.S. 224, 240, 93 S.Ct. 810, 818, 35 L.Ed.2d 223, 236 (1973). Petitioners were afforded a meaningful opportunity to be heard and to controvert the evidence. Fairness demands no more. See International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 427, 478 F.2d 615, 631 (1973).
PPG, duPont and NPRA claim the Administrator was required to set ambient air standards for lead under § 108, 42 U.S.C. § 1857c-3, before acting under § 211, and his failure to do so warrants reversal. PPG/duPont Br. at 45; NPRA Supp.Br. at 37-47. There is no basis in the statute for this claim. While § 211 expressly mentions consideration of regulation under § 202, it makes no mention at all of § 108. Moreover, the Administrator claims, with some statutory support, that action under § 108 is discretionary with him. 38 Fed.Reg. 33740 (1973). Cf. note 21 supra.
Nalco, PPG and duPont all claim that the Administrator did not give sufficient consideration to the adverse economic effects of his decision to regulate lead additives. Nalco Br. at 42; PPG Br. at 40. Even if the Administrator is allowed to consider at all the economic effects of regulations issued under § 211(c)(1)(A), a question we do not address, he has plainly both well understood and considered the consequences of these regulations and found them to be minimal. 38 Fed.Reg. 33739 (1973). We cannot say his judgment was irrational.
NPRA objects to the Administrator’s switch from leaded pool averaging to total pool averaging. NPRA Supp.Br. at 47-48. EPA switched in response to the comments of a majority of refiners. 38 Fed.Reg. 33739. We find nothing wrong with the use of that system. We discussed at note 100 supra the propriety of the Agency’s announcing this change when promulgating its final regulations.
Lastly, Ethyl, PPG, duPont and NPRA claim that EPA was required to file an environmental impact statement under § 102(2)(C) of the National Environmental Policy Act, 42 U.S.C. § 4332(2)(C) (1970), or.the functional equivalent of one. Ethyl Br. at 57; PPG Br. at 46; NPRA Br. at 53. So far as EPA regulation under § 211 is concerned, we answered this question in Amoco, and recent legislation has made the point doubly clear. An environmental impact statement from EPA does not appear necessary; a “functional equivalent” suffices. The Amoco court held that when the Administrator has considered the evidence and alternative courses of action required before action under § 211 is permitted, he has necessarily furnished the “functional equivalent” of an impact statement. Amoco Oil Co. v. EPA, supra note 2, 163 U.S.App.D.C. at 189-190, 501 F.2d at 749-750. Since we believe the Administrator has complied with all the demands of § 211 in this case, we believe he has likewise complied with NEPA.
Any lingering doubt in this regard has been removed by recent congressional action. The Energy Supply and Environmental Coordination Act of 1974, Pub.L.No.93-319, 88 Stat. 246 (1974) (codified at 15 U.S.C. § 791 et seq. (Supp. IV 1974)), provides: “No action taken under the Clean Air Act shall be deemed a major Federal action significantly affecting the quality of the human environment within the meaning of the National Environmental Policy Act of 1969.” Id. § 7(c)(1), 15 U.S.C. § 793(c)(1). Thus Congress has expressly exempted EPA action under the Clean Air Act from NEPA’s environmental impact statement requirement.
. R. Kehoe, The Metabolism of Lead in Man in Health and Disease: The Harben Lectures (1960), reprinted from J. Royal Inst. Public Health Hyg. 1961, JA 500.
. The absorption data is based on an estimated 15 m8 daily respiration (since the subjects were sedentary), 30-37% absorption of lead from air, and a lead concentration of 10.9 ug/m 8. Third Health Document at Table V-5, JA 104.
. Petitioners’ second objection is likewise without merit. Nalco claims EPA improperly assumes a straight-line correlation between air lead exposure and blood lead levels, and quotes EPA scientists:
The best fit for these [Kehoe] data may be curvilinear rather than a linear function, in which case, blood leads would level off about 40 ug/100 g rather than continuing to increase indefinitely.
JA 85, cited in Nalco Supp.Br. at 39. The EPA calculations, however, make just the adjustment suggested (which is hardly surprising since the suggestion appears in the Third Health Document). EPA calculates only the air lead levels necessary, under various circumstances, to bring blood leads to 40 ug — and no higher. Whether or not blood lead increases plateau thereafter is of no consequence for these calculations.
. Climate may influence both air lead levels and exposure to lead. Greater wind speed is associated with lower air lead levels, and higher readings have been associated with surface inversions. JA 854, 873. Climate also influences the amount of time spent exposed to atmospheric lead outdoors, the time spent indoors with windows open (which increases lead exposure) or air conditioners on (which decreases lead exposure). JA 400, 467. See note 11 infra. For these reasons, it may be preferable to compare blood lead levels of persons exposed to roughly similar climates.
. PPG and duPont argue that this conclusion is contradicted by the Philadelphia data in the Seven Cities Study. Philadelphia urban-suburban data were analyzed twice, they argue, and the second analysis turned the result around, producing higher blood lead values for suburban dwellers than for city dwellers. PPG/du-Pont Supp.Br. at 25. In fact, the data were analyzed three times, the first time in the Seven Cities Study itself, JA 901 (Ardmore and Rittenhouse data), and the second and third times in a submission prepared for EPA by duPont. Only on the third analysis did the turn-about result. JA 2211. Thus two out of three analyses of the Philadelphia data (one by duPont), as well as the data for the other two cities, support the Administrator’s position. We cannot say his conclusion is irrational.
. Azar et al., Relationship of Community Levels of Air Lead and Indices of Lead Absorptions, in Proceedings of an International Symposium on Environmental Health Aspects of Lead, CEC, CID, Luxembourg, at 581-594 (May 1973), JA 387.
. Nalco attempts to belittle this finding by claiming that it represents “a comparison of 6 cab drivers with 2 office workers A. 392).” Nalco Supp.Br. at 35-36. Nalco is distorting the record. The cited reference, Table 1 of the Azar study, JA 392, is explicitly headed “30 subjects per group.” What Nalco is apparently referring to is a line reading, “Data Per Subject, # Bloods: L.A. Cab, 6; L.A. Office, 2 * * Contrary to Nalco’s suggested reading, this line indicates only that the 30 taxicab drivers were given 6 blood tests each, while the 30 office workers were given only 2 blood tests each.
.Tsuchiya et al., Study of Lead Concentrations in Atmosphere and Population in Japan (undated), JA 1092.
. Jones et al., Blood Lead and Carboxyhemoglobin Levels in London Taxi Drivers, 2 Lancet 302 (Aug. 12, 1972), JA 498.
. Daines et al., Air Levels of Lead Inside and Outside Homes, 41 Industrial Med.J. 26 (Oct. 1972), JA 465.
. Among those women living 3.7 meters from the highway, both air lead and blood lead levels were significantly lower in homes that were air-conditioned. JA 467; Third Health Document at V-10, JA 90. See note 4 supra.
. In both studies air lead concentrations were comparable and blood lead differences were insignificant at distances beyond 100 feet from the roadway. In the first study, at distances of 38.1 and 121.9 meters the air lead concentrations were 2.41 and 2.24 ug/m3 and blood lead concentrations 17.4 and 17.6 ug. In the second study, at distances of 33.5 and 457 meters the air lead concentrations were 1.95 and 1.73 ug/m3 and the blood lead concentrations 15.7 and 16.1 ug. All of this shows only that where air lead concentrations are essentially similar, there is no reason to anticipate differences in blood lead concentrations.
. Hammer et a 1., “Trace Metals in Human Hair as a Simple Epidemiologic Monitor of Environmental Exposure,” Trace Substances in Environmental Health V; A Symposium 25 (D. Hemphill, ed. 1972).
. Nordman, Blood Lead Levels and Erythrocyte Delta-Aminolevulinic Acid Dehydratase Activity in People Living Around a Secondary Lead Smelter, 10 Work — Environment—Health 19 (1973).
. Petitioners accuse the Administrator of worrying about dietary lead intake only when it serves his purpose of discrediting a contradictory study, and conveniently ignoring the problem when he wants to use the data. Ethyl Supp.Br. at 52. In fact, however, the Administrator’s concern is consistent. Dietary control is a problem in all the epidemiological studies. The Administrator consistently credits inferences derived from data in a single metropolitan area or where dietary absorption of lead can be deemed roughly constant. This is a principled approach to the evidence, and one which we approve.
. Knelson et al, Kinetics of Respiratory Lead Intake in Humans, in Proceedings of an International Symposium on Environmental Health Aspects of Lead, CEC, CID, Luxembourg, at 391-401 (May 1973), JA 596.
. Rabinowitz et ai., “Study of human lead metabolism using stable isotope tracers,” paper presented at EPA-NIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C., Oct. 1-2, 1973, JA 678; Rabinowitz et al., “Lead Metabolism in the Normal Human; Stable Isotope Studies,” 182 Science 725 (Nov.1973), JA 704. See Third Health Document at V-3, JA 83. See also 38 Fed.Reg. at 33735.
. See, e. g., NYC Bureau of Lead Poisoning Control, data submitted to EPA Aug. 31, 1972 & Sept. 12, 1972, Doc. 104; Lepow, testimony before Subcommittee on the Environment, Senate Committee on Commerce, May 8, 1972, cited in Second Health Document at V-18, JA 196.
. Personal communication from M. Dea, Department of Human Services, Multnomah County, Oregon, to K. Bridbord, EPA, dated March 6, 1973, Doc. 44; Gilsinn, Estimates of the Nature and Extent of Lead Paint Poisoning in the United States, NBS Technical Note 746, Dec. 1972, Doc. 61; Needleman & Shapiro, “Dentine Lead Levels in Asymptomatic Philadelphia School Children: Subclinical Exposure in High and Low Risk Groups,” presented at EPA-NIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C. (Oct. 1-2, 1973), JA 662.
. Sayre et a 1., House and Hand Dust as a Potential Source of Childhood Lead Exposure, —• Am.J.Dis.Child — (1973) (in press at time of insertion in record), JA 720; Vostal et al., “Lead Analysis of the House Dust: A Method for the Detection of Another Source of Lead Exposure in Inner City Children,” presented at EPA-NIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C. (Oct. 1-2, 1973), JA 738.
Since only 58% of the inner-city households had lead-based paint, lead paint chips were discounted as the sole source of elevated lead levels. The study made no attempt to determine “whether the lead comes from airborne sources, chalking of paint, or from paint chips shed from walls and perhaps pulverized underfoot.” JA 732.
.The authors concluded, “[M]ost cases of pediatric Pb [lead] poisoning occur in an area of high soil Pb values.” Fairey & Gray, Soil Lead and Pediatric Lead Poisoning in Charleston, S.C., J.S.C. Med. Assn 79 (March 1970), JA 482, 485. Petitioners distort the purpose of this study and then attack the resultant straw man. As petitioners correctly note, Nalco Supp.Br. at 43-44; PPG/duPont Supp.Br. at 43; Ethyl Supp.Br. at 50, the high lead concentrations in the area studied were not thought to be a direct product of automotive exhausts. See JA 484, 485. Since the study is only cited to show that children are likely to eat lead-contaminated dust, however, petitioners’ charge proves little. Third Health Document at VI-4, JA 114. On the other hand, petitioners fail to note that the study did implicate automotive exhausts to some extent:
Neither pesticides nor automobile exhausts account for the high Pb [lead] levels in Charleston, though undoubtedly they add to the contamination.
JA 484.
Petitioners attempt to rebut the Charleston findings, and those of the El Paso study, see note 5 infra and accompanying text, by relying on apparently contradictory data from England showing inconsequential increases in children’s blood lead levels despite significant changes in soil lead levels. PPG/duPont Supp.Br. at 43-44; Ethyl Br. at 42, citing Barltrop & Strehlow, “The Significance of High Soil Lead Concentrations for Childhood Lead Burdens,” presented at EPA-NIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C. (Oct. 1-2, 1973), JA 415. *436But see EPA Supp.Br. at 57 n.51. The Administrator pointed to various possible differences between American and English data that might account for the inconsistency. Third Health Document at VI-5, JA 115. We shall defer to his judgment that the American data are preferable.
. Within one mile of the smelter 35.2% of the children tested had blood lead levels of 40 ug or greater; beyond that the percentage dropped to around 10. Epidemiological Studies of Human and Environmental Lead Pollution: El Paso, Texas (unpublished 1972), JA 468 (Table 2). On the basis of this data the chairman of the NAS Panel concluded;
[Tjhe Smeltertown [El Paso] study alone lends a high degree of credibility to the hypothesis that very young children will, in fact, ingest toxicologically significant amounts of dust and dirt when it contains lead at the level of 0.1 percent and up.
Letter from Dr. Paul B. Hammond to Dr. Kenneth Bridbord, EPA, March 9, 1973, JA 2417. While petitioners cited an apparently contradictory statement from Dr. Hammond, JA 2611, he has not repudiated the observation quoted above.
. Margulis et al.t “Residential Location, Ambient Air Lead Pollution and Childhood Lead Poisoning,” — Arch. Env. Health — (in press at time record was assembled), JA 626. Petitioners suggest that something is wrong with the Newark data because preliminary results did not reach the same conclusion. Nalco Supp.Br. at 42-43; PPG/duPont Supp.Br. at 40-41. Preliminary calculations on data points only above lead levels of 40 ug produced positive, but not significant, results, so the data field was expanded to include all the evidence whereby the results suggested below were obtained. We see nothing wrong with such refinement of experimental data. See also pp. 422-423 n.112 of 176 U.S.App.D.C., pp. 50-51 n.112 of 541 F.2d supra.
. Over 5,000 children were studied. Of those living within 100 feet of major highways, 49.3% had blood lead levels of 40 — 49 ug and 8.1% had levels of 60 ug or more; of children living between 100 and 200 feet from the highways, the respective percentages were 24.2 and 3.7. Margulis et al., supra note 6, at Table 2, JA 639. Dividing these children into two groups, one containing children living near highways bearing an average weekday traffic density of less than 24,000 vehicles, the other containing children living near highways bearing more than 24,000 vehicles daily, the researchers found 51.3% of the latter group had blood lead levels of 40-59 ug, and 10.8% had levels greater than 60 ug; in the former group, however, the respective percentages were only 3.6 and 5.1.
Undoubtedly, respiration of airborne lead emissions contributed to the elevated blood lead levels, but the fact that a far greater percentage of these children had elevated blood lead than comparable adults, see the Daines study discussed at pp. 431 to 432 of 176 U.S.App.D.C., pp. 59-60 of 541 F.2d supra, suggests that ingestion of lead dustfall significantly supplemented any respiratory intake.
. In particular, petitioners fault the Newark study for failing to correlate the children’s blood lead level with exposure to leaded paint as well as distance from the roadway. Nalco Supp.Br. at 43; PPG/duPont Supp.Br. at 41-42; Ethyl Supp.Br. at 51. These data were impossible to obtain; the Newark study took advantage of a vast data base accumulated by the New Jersey College of Medicine and Dentistry in its routine measurements of childhood exposure to lead. The data contained only blood lead levels and addresses of children, thus making the lead paint variable impossible retroactively to control or measure, JA 954. However, the Administrator recognized this limitation on the study, Third Health Document *437at VI-17, JA 127, and determined that it did not seriously undercut its validity. Of course, the results were “only” a statistical correlation; such are the results of all epidemiological research. But the Newark study presented a strong correlation, consistent with an existing theory that makes biological sense and with existing evidence. Such data are worthy of great respect, and may even be taken as causal proof. H. Hilleboe & G. Larimore, Preventive Medicine 677-678, 687 (1959). We cannot say the Administrator’s reliance on the Newark study was in error.
Petitioners also argue the El Paso data is useless since blood lead levels were elevated in adults as well, so that the high lead body burdens may have been due to inhalation rather than ingestion. Nalco Supp.Br. at 41-42. While this suggestion may undercut petitioners’ argument that inhalation of airborne lead does not significantly affect the lead body burden, it does not undercut the Administrator’s use of the study. While adult blood leads were indeed elevated, the Administrator noted that the frequency of elevated blood leads among young children was much higher, thus suggesting “a special exposure route for young children,” i. e., ingestion. Third Health Document at VI-7, JA 117. (Petitioners overstate by at least a factor of 10 the air lead concentrations in El Paso, compare Nalco Supp.Br. at 41 with JA 117, 469, 2344.)
. Ter Haar & Aronow, New Information on Lead in Dirt and Dust as Related to the Childhood Lead Problem, presented at the EPANIEHS Conference on Low Level Lead Toxicity, Raleigh, N.C., Oct. 1, 1973, JA 917, 922 (Fig. 1); Fine et a/., Pediatric Biood Lead Levels, 221 J. A.M.A. 1475 (1972), JA 486; Conn & Anderson, D.C. Mounts Unfunded Program of Screening for Lead Poisoning, 86 H.S.M.H.A.. Health Reports 409 (1971), JA 460; statement of Henrietta Sachs (Nov. 5, 1972), JA 2444.
. Ter Haar & Aronow, supra note 9.
. As the study concluded: “The results provide sound evidence that these children suspected of elevated lead body burden were not ingesting dust or air-suspended particulate.” JA 918 (emphasis added).